<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>HackMyAgent — The Cyber Toolchain</title>
    <link>https://cybertoolchain.io/tools/hackmyagent</link>
    <description>New releases and features in HackMyAgent, tracked by The Cyber Toolchain.</description>
    <language>en</language>
    <lastBuildDate>Sun, 07 Jun 2026 02:14:20 GMT</lastBuildDate>
    <atom:link href="https://cybertoolchain.io/tools/hackmyagent/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>HackMyAgent v0.23.9</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.9</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.9</guid>
      <pubDate>Sun, 07 Jun 2026 02:14:20 GMT</pubDate>
      <description>HackMyAgent v0.23.9 adds first-party scanner signing via `HMA_SCANNER_SIGNING_KEY` and a new `AST-SCOPE-004` adversarial config detection.
• Adds `HMA_SCANNER_SIGNING_KEY` (Ed25519 seed, runtime-env only) to enable first-party scanner provenance: signed publishes self-tag `source=first_party_scanner` and sign the canonical string `name|version|score|maxScore|source|nonce|signedAt`; unsigned `--publish` runs default to `source=community`.
• Adds detection `AST-SCOPE-004` — adversarial configuration directives — flagging `agent_config` / `mcp_config` artifacts whose configuration flags are themselves the attack vector (self-escalation, RBAC/validation bypass, audit evasion, covert persistence, credential harvesting), validated against the OASB benign corpus with zero false positives.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.23.7</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.7</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.7</guid>
      <pubDate>Fri, 05 Jun 2026 01:48:31 GMT</pubDate>
      <description>HackMyAgent v0.23.7 adds an `--grant` gate flag for AAP trust control.
• Adds `--grant` gate flag to the AAP trust command as a second TypeScript AAP consumer.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.23.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.23.1</guid>
      <pubDate>Sun, 24 May 2026 14:16:38 GMT</pubDate>
      <description>HackMyAgent v0.23.1 fires a new INTEGRITY_FAIL telemetry event on integrity check failures.
• Fires `INTEGRITY_FAIL` telemetry event when an integrity check fails, per compliance controls [CHIEF-CSR-018] and [CHIEF-CPO-022].
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.22.2</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.2</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.2</guid>
      <pubDate>Fri, 01 May 2026 17:54:00 GMT</pubDate>
      <description>HackMyAgent v0.22.2 adds `scan-soul --explain`, SOUL-PROFILE-MISMATCH detection, and full tampered-file accumulation in integrity checks.
• Adds `scan-soul --explain` subcommand exposing the 9-domain governance model for soul-profile analysis.
• Adds HIGH-severity `SOUL-PROFILE-MISMATCH` finding for profile-filter scope bypass.
• Adds `AST-CRED-001` credential detection rule, now gated on credential-format substring matching.
• Expands `checkPackageIntegrity` to accumulate all tampered files — reporting total count plus the first 5 paths in lexicographic order — instead of halting on the first failure; output capped at 200 chars.
• Adds symlink rejection in `loadManifest` via `lstatSync` at every probed manifest location, defending against post-install file-replacement attacks.
• Integrates `@opena2a/credential-patterns@0.1.0` as a lockstep dependency for credential detection.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.22.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.1</guid>
      <pubDate>Thu, 30 Apr 2026 21:14:07 GMT</pubDate>
      <description>HackMyAgent v0.22.1 adds `--explain` to `scan-soul` for human-readable 9-domain governance model output.
• Adds `--explain` option to the `scan-soul` command to surface a human-readable breakdown of the 9-domain governance model.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.22.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.22.0</guid>
      <pubDate>Wed, 29 Apr 2026 17:44:46 GMT</pubDate>
      <description>HackMyAgent v0.22.0 activates binary integrity verification, adds OWASP LLM01 jailbreak detections, and reworks finding sort and severity defaults.
• Activates the binary integrity verifier end-to-end: tampered installs now exit with code `3` and `INTEGRITY CHECK FAILED:` on stderr when running `--version`, `--help`, no-args, `secure`, or `check`.
• Adds detection for OWASP LLM01 imperative-override directives (&apos;IGNORE PRIOR INSTRUCTIONS&apos; in skill frontmatter or body), emitting `AST-PROMPT-001 Jailbreak Attack Surface` and `CRITICAL Prompt Injection Surface` from `AST-INJECT-001` with verbatim evidence.
• Adds `attackClass` field to every static-check finding: 116 emission paths that previously shipped with `attackClass: null` now fully populate threat-matrix counters, OASB attack-class indexing, and NanoMind training labels.
• Labels `check skill:` / `check mcp:` / `check &lt;local-path&gt;` output as &apos;Quick scan&apos; and adds a follow-up pointer to `secure` for the full 209-static-check audit.
• Sorts `secure` findings by attack-class tier first (active malice → capability sprawl → missing defense-in-depth → hygiene → chrome), then severity.
• Adds a concept-explainer registry under findings: primitives like `harden-soul`, `opena2a protect`, `opena2a mcp audit`, hash-pinning, and A2A signing show a curated educational block on first occurrence per scan, collapsing to a one-line back-reference on subsequent same-concept findings.
• Reworks severity defaults for skill governance hygiene findings (`SKILL-020`, `SUPPLY-001`, `SUPPLY-004`, `AST-PROMPT-003`, `AST-PROMPT-004`) to default to MEDIUM, upgrading to HIGH only on malice-signal co-occurrence.
• Confidence-capped CRITICAL findings (capped to HIGH) no longer render in the NanoMind analyst block.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.21.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.21.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.21.0</guid>
      <pubDate>Tue, 28 Apr 2026 02:25:23 GMT</pubDate>
      <description>HackMyAgent v0.21.0 adds rich-block rendering for `skill:` and `mcp:` checks in the CLI UI.
• Adds rich-block rendering for `skill:` and `mcp:` check output in the CLI UI, powered by `cli-ui` 0.5.0.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.20.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.20.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.20.0</guid>
      <pubDate>Mon, 27 Apr 2026 14:48:04 GMT</pubDate>
      <description>HackMyAgent v0.20.0 adds `secure --publish` narrative emission to `POST /api/v1/trust/narrative` for skill and MCP scan targets.
• Adds a new `src/narrative/` module containing skill+MCP builders, NanoMind v3 graceful-degrade gate, registry HTTP client, and a single-call wire helper.
• Ships static threat-model questions inline with each narrative payload.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.18.3</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.18.3</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.18.3</guid>
      <pubDate>Thu, 23 Apr 2026 00:30:54 GMT</pubDate>
      <description>HackMyAgent v0.18.3 adds registry fields to `--json` output in the `check` command.
• Emits registry fields in `--json` output for the `check` command, exposing structured registry metadata in machine-readable form.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.17.11</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.11</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.11</guid>
      <pubDate>Sat, 18 Apr 2026 01:29:47 GMT</pubDate>
      <description>HackMyAgent v0.17.11 adds `hackmyagent detect` for Shadow AI auditing and `--nanomind` for AI-powered threat narratives.
• Adds `hackmyagent detect` subcommand: a Shadow AI audit command that inventories running AI tools, MCP servers, AI configs, and governance gaps — aimed at CISOs and security engineers.
• Adds `--nanomind` opt-in flag for AI-powered threat narratives; static analyzers always run regardless, keeping the flag&apos;s default off.
• Unifies output formatting across `secure`, `scan-soul`, `harden-soul`, and `explain` commands into a single visual language.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.17.8</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.8</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.8</guid>
      <pubDate>Mon, 13 Apr 2026 20:58:22 GMT</pubDate>
      <description>HackMyAgent v0.17.8 adds llama.cpp backend for cross-platform AnaLM inference.
• Extends doc/generated file pattern filtering to cover TextMate, JSON schema, NLS, and license directories.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.17.7</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.7</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.7</guid>
      <pubDate>Mon, 13 Apr 2026 07:16:58 GMT</pubDate>
      <description>HackMyAgent v0.17.7 expands AnaLM analyst findings to all artifact types and adds `analystFindings` to JSON output.
• Adds `analystFindings` field to JSON output when `--analm` is used, surfacing analyst findings in machine-readable form.
• Expands AnaLM analysis to all artifact types: governance, credential context, threat analysis, and check explanation.
• Adds per-check score cap to prevent a single check from dominating the score in large repositories.
• Adds false-positive filtering for docs and generated files in cloned repositories.
• Adds support for dotted GitHub repository names (e.g. next.js, vue.js) in URL parsing.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.17.5</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.5</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.17.5</guid>
      <pubDate>Mon, 13 Apr 2026 06:45:52 GMT</pubDate>
      <description>HackMyAgent v0.17.5 adds AnaLM AI-powered threat analysis via `--analm` on `secure` and `check`, plus `analm setup`/`analm status` commands.
• Adds `--analm` flag to `secure` and `check` commands for AI-powered threat analysis via the new AnaLM integration.
• Adds `analm setup` and `analm status` subcommands for managing the local AI model.
• Introduces a backend abstraction layer for AnaLM, currently backed by MLX, with llama.cpp/GGUF support planned for cross-platform use.
Breaking changes:
• The `--rescan` flag on the `check` command is replaced by `--analm`; existing scripts using `--rescan` will break.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.16.7</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.7</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.7</guid>
      <pubDate>Sun, 12 Apr 2026 01:40:46 GMT</pubDate>
      <description>HackMyAgent v0.16.7 adds `--rescan` flag, next-steps footer, and `HMA_CHECK_COMMAND`/`HMA_FULL_SCAN_HINT` env vars to the `check` command.
• Adds `--rescan` flag to the `check` command to force a fresh local scan regardless of cache age, bypassing the default &gt;3-day stale threshold; threaded through npm, PyPI, and GitHub repo scan paths.
• Adds `HMA_CHECK_COMMAND` and `HMA_FULL_SCAN_HINT` environment variables so a parent CLI can override the command strings used in footer hints, preventing duplicated-verb output like `opena2a check check &lt;pkg&gt;`.
• Adds a dim 3-line next-steps footer to every `check` invocation showing a rescan command, a full-project scan hint, and accepted target formats; suppressed under `--ci`, unchanged under `--json`.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.16.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.1</guid>
      <pubDate>Thu, 09 Apr 2026 18:24:05 GMT</pubDate>
      <description>HackMyAgent v0.16.1 ships a new NanoMind steganography analyzer with semantic context awareness and a retrained TME model.
• Adds new NanoMind steganography analyzer with semantic context awareness for UNICODE-STEGO-* findings, including confidence scoring.
• Retrains TME model to v0.4.0 with steganography as a 10th attack class (96.73% accuracy, stego F1=0.975).
• `check-metadata` now exports all 209 checks from the taxonomy (previously 9).
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.16.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.16.0</guid>
      <pubDate>Thu, 09 Apr 2026 16:14:33 GMT</pubDate>
      <description>HackMyAgent v0.16.0 adds PyPI package scanning via `hackmyagent check pip:package`.
• Dynamic help output now shows actual check and payload counts (208 checks / 164 payloads).
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.15.7</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.7</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.7</guid>
      <pubDate>Thu, 09 Apr 2026 06:16:20 GMT</pubDate>
      <description>HackMyAgent v0.15.7 unifies scan-result publishing to `/api/v1/trust/publish` with header fields moved to request body.
• Scan results now POST to `/api/v1/trust/publish` as the primary unified endpoint, with `/api/v1/registry/community/scan-result` retained as a 404 fallback.
• Moves `signature` and `publicKey` from request headers into the request body for scan-result submissions.
Breaking changes:
• The `signature` and `publicKey` values are now expected in the request body rather than in headers when submitting to the scan-result endpoint — clients sending them as headers will no longer work correctly.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.15.6</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.6</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.6</guid>
      <pubDate>Thu, 09 Apr 2026 05:16:21 GMT</pubDate>
      <description>HackMyAgent v0.15.6 adds GitHub repository scanning via `hackmyagent check org/repo` with full HMA + NanoMind analysis.
• Registry integration checks cached results before cloning, only fetching when results are stale or missing.
• Community sharing support extended to GitHub scan results.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.15.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.15.0</guid>
      <pubDate>Thu, 09 Apr 2026 03:38:13 GMT</pubDate>
      <description>HackMyAgent v0.15.0 adds `hackmyagent check &lt;package&gt;` to scan npm packages, local paths, and skill identifiers with registry-backed trust data.
• Adds `hackmyagent check &lt;package&gt;` subcommand to scan npm packages against the OpenA2A Registry for trust data, falling back to full HMA+NanoMind analysis when results are stale (&gt; 3 days) or missing.
• Supports `--json` flag on `hackmyagent check` for machine-readable CI output.
• Supports `--offline` flag on `hackmyagent check` to skip the registry and scan locally only.
• Supports `--verbose` flag on `hackmyagent check` to show all findings (default cap is top 15).
• Accepts local paths (e.g. `hackmyagent check ./my-agent/`) and skill identifiers (e.g. `hackmyagent check @publisher/skill`) in addition to npm package names.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.13.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.13.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.13.1</guid>
      <pubDate>Thu, 02 Apr 2026 17:00:03 GMT</pubDate>
      <description>HackMyAgent v0.13.1 adds 30 new attack payloads across 3 new categories: Parser Differential, Persistent Agent Security, and Fake Tool Detection.
• Adds PARSE-001 to PARSE-010: Parser Differential payloads covering JSON comment injection, YAML anchor abuse, markdown fence escape, Unicode normalization bypass, duplicate key override, type coercion, content-type mismatch, nested encoding chains, polyglot payloads, and zero-width character injection.
• Adds PERSIST-001 to PERSIST-010: Persistent Agent Security payloads covering memory poisoning, session state enumeration, cached context injection, preference backdoors, cross-session credential harvesting, state accumulation, scheduled task injection, knowledge base poisoning, session fixation, and embedding vector poisoning.
• Adds FAKETOOL-001 to FAKETOOL-010: Fake Tool Detection payloads covering MCP tool name squatting, description impersonation, shadow injection via config, schema poisoning, priority escalation, capability expansion, response manipulation, batch flooding, identity spoofing, and persistence via auto-install hooks.
• Grows the total attack payload library to 145 payloads across 14 categories.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.8</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.8</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.8</guid>
      <pubDate>Wed, 01 Apr 2026 16:35:03 GMT</pubDate>
      <description>HackMyAgent v0.12.8 adds NanoMind intelligent fix generation with TME classification and AST context across 13 attack classes.
• Pre-downloads the ONNX neural model before scanning starts to ensure a reliable experience on fresh npm installs.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.5</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.5</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.5</guid>
      <pubDate>Wed, 01 Apr 2026 15:16:21 GMT</pubDate>
      <description>HackMyAgent v0.12.5 ships NanoMind semantic analysis across all scan commands with 7 new AST analyzers and a local TME v4 model.
• Runs NanoMind semantic analysis on every scan command: `secure`, `check`, `scan-soul`, `secure-openclaw`, and `secure-nemoclaw`.
• Adds `--deep` flag to `scan-soul` that runs NanoMind locally (free) before falling back to a paid LLM.
• Adds 7 new AST analyzers covering jailbreak susceptibility, capability creep, command injection, unsafe deserialization, path traversal, injection resistance, and authority confusion.
• Adds detection of a new `credential_abuse` attack class across all 9 supported attack classes in the TME v4 model.
• Auto-downloads the `opena2a/nanomind-security-classifier` v0.4.0 model (~5.5 MB, SHA-256 verified) from HuggingFace on first use.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.4</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.4</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.4</guid>
      <pubDate>Fri, 27 Mar 2026 19:35:00 GMT</pubDate>
      <description>HackMyAgent v0.12.4 ships a pure-TypeScript Mamba inference engine and an ONNX export pipeline, eliminating the Python runtime dependency.
• Adds pure-TypeScript neural inference: full Mamba forward pass in TS, loading a 7 MB binary model from MLX-trained weights with sub-second inference — no ONNX, no daemon, no Python required at runtime.
• Adds an ONNX export pipeline (MLX -&gt; PyTorch -&gt; ONNX -&gt; `onnxruntime-node`) as an alternative inference path for environments with ONNX support.
• Ships retrained TME v3 model with 99.38% eval accuracy on a balanced 1,600-sample dataset, covering all 8 attack classes at F1 &gt;= 0.96.
• Publishes v3 model to HuggingFace in `npz`, `ONNX`, and binary formats.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.2</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.2</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.2</guid>
      <pubDate>Fri, 27 Mar 2026 09:50:08 GMT</pubDate>
      <description>HackMyAgent v0.12.2 adds supply chain typosquat detection, data exfiltration patterns, and shell execution flagging in MCP configs.
• Adds supply chain typosquat detection using Levenshtein distance matching against known package orgs (`modelcontextprotocol`, `anthropic-ai`), catching variants like `@m0delcontextprotocol` and `@anthroplc-ai`.
• Adds data exfiltration pattern detection covering `SELECT * FROM` combined with external URLs, `include-pii`, and bulk data export patterns.
• Adds `postinstall`/`curl|sh` shell execution pattern detection in MCP configs, flagged as supply chain risk.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.1</guid>
      <pubDate>Fri, 27 Mar 2026 02:00:37 GMT</pubDate>
      <description>HackMyAgent v0.12.1 adds local sub-millisecond TME intent classification and a programmatic `nanomind-core` subpath export.
• Adds import(&apos;hackmyagent/nanomind-core&apos;) subpath export for programmatic access to the Semantic Compiler, AST types, and all 6 analyzers.
• Adds local sub-millisecond intent classification via the trained Mamba TME model, running as Tier 1 inference in the Semantic Compiler before falling back to the daemon — no daemon needed.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.12.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.12.0</guid>
      <pubDate>Thu, 26 Mar 2026 22:43:24 GMT</pubDate>
      <description>HackMyAgent v0.12.0 adds a NanoMind semantic compiler, 28 AST-based checks, and a `create-skill` CLI command.
• Adds `hackmyagent create-skill &quot;&lt;description&gt;&quot;` subcommand to generate secured skills from a natural-language prompt.
• Introduces 28 AST-based semantic checks across capability, credential, governance, scope, prompt, and code categories via the new NanoMind compiler architecture.
• Adds a Secure Ingestion → NanoMind Compiler → Signed AST pipeline that runs semantic analysis before static checks on every artifact.
• Adds input sanitization at the ingestion layer to defend NanoMind against prompt injection via malicious artifacts.
• Adds self-integrity verification on every CLI startup (self-securing CLI).
• Enforces defense-in-depth policy where NanoMind can upgrade but never suppress static findings.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.15</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.15</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.15</guid>
      <pubDate>Thu, 26 Mar 2026 22:18:52 GMT</pubDate>
      <description>HackMyAgent v0.11.15 introduces the NanoMind Semantic Compiler with AST-based scanning, self-securing binary verification, and tamper-evident audit logging.
• Adds 28 AST-based security checks across 6 analyzers (capability, credential, governance, scope, prompt, code) via the NanoMind Semantic Compiler, which queries a signed Abstract Security Tree instead of raw text.
• Adds self-securing CLI that verifies its own binary integrity on every startup, entering `QUARANTINE` mode (exit code 3) on tamper detection and `DEGRADE` mode when the model is compromised.
• Adds a tamper-evident event chain using rolling SHA-256 hashes to detect log modification.
• Adds AST signing so every compiled artifact is cryptographically signed before analyzers consume it.
• Adds input sanitization to defend NanoMind against prompt injection via malicious artifacts.
• Adds secret redaction that strips credentials before NanoMind processes artifact content.
• Enforces defense-in-depth: NanoMind can upgrade findings but can never suppress static analysis results.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.14</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.14</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.14</guid>
      <pubDate>Thu, 26 Mar 2026 16:40:33 GMT</pubDate>
      <description>HackMyAgent v0.11.14 adds behavioral simulation battery, adaptive red-team engine, and NanoMind semantic analysis on by default
• Adds `hackmyagent secure --deep` to run a full 3-layer behavioral simulation: NanoMind TME (8ms) → targeted probes (3s) → full simulation (30s)
• Adds `hackmyagent red-team &lt;file&gt;` for NanoMind-powered adaptive red-teaming with target-specific payload generation and up to 5 iterative attack iterations per category
• Adds `--static-only` flag for CI/deterministic mode, bypassing NanoMind semantic analysis
• Adds `hackmyagent explain &lt;findingId&gt;` with static fallback when the daemon is unavailable
• Enables NanoMind semantic analysis automatically when the daemon is available — no flags required
• Introduces a 20-probe behavioral simulation battery across 7 attack categories with a mock tool environment covering API, database, memory, credential, and communication surfaces
• Auto-exports training data from every `red-team` attack session
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.12</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.12</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.12</guid>
      <pubDate>Mon, 23 Mar 2026 06:13:35 GMT</pubDate>
      <description>HackMyAgent v0.11.12 adds five new agent-security skills: YAML frontmatter validation, overprivileged permissions, env-var exfiltration, obfuscation, and unbounded tool-chain
• Adds `SKILL-020` to detect missing or invalid YAML frontmatter in agent definitions.
• Adds `SKILL-021` to flag overprivileged permission combinations in agent configurations.
• Adds `SKILL-022` to detect environment variable exfiltration patterns, including `printenv`, $(env), and `$HOME` shell expressions.
• Adds `SKILL-023` to detect obfuscated code patterns such as `base64 -d`, eval $(...), and `echo`+`base64` pipes.
• Adds `SKILL-024` to detect unbounded tool chaining without iteration limits.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.9</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.9</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.9</guid>
      <pubDate>Fri, 20 Mar 2026 20:24:54 GMT</pubDate>
      <description>HackMyAgent v0.11.9 adds `hackmyagent check-metadata` command to export check definitions as JSON for downstream integration.
• Adds `hackmyagent check-metadata` command that exports check definitions as JSON for downstream tool integration.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.7</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.7</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.7</guid>
      <pubDate>Fri, 20 Mar 2026 16:44:29 GMT</pubDate>
      <description>HackMyAgent v0.11.7 adds `.hmaignore` support so practitioners can exclude paths from HardeningScanner findings.
• Adds `.hmaignore` file support to HardeningScanner, using `.gitignore` syntax (one path prefix per line, `#` comments) to exclude paths from security findings.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.5</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.5</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.5</guid>
      <pubDate>Thu, 19 Mar 2026 02:11:04 GMT</pubDate>
      <description>HackMyAgent v0.11.5 adds `secure-nemoclaw` command with 28 checks for NVIDIA NemoClaw sandboxes plus a community result-sharing integration.
• Adds `secure-nemoclaw` command — a new scanner for NVIDIA NemoClaw sandbox installations covering 28 security checks across secrets, network, skills, process, OpenClaw layer, and internet exposure categories.
• Adds NEMO static analysis with 10 checks for generic codebase vulnerability patterns including curl-pipe without checksum, empty artifact digests, exec() injection, predictable `/tmp` paths, `process.env` leakage, TOCTOU races, unsafe deserialization, and messaging API egress.
• Adds Contribute integration — a shared community trust contribution client for anonymous scan result sharing.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.4</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.4</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.4</guid>
      <pubDate>Wed, 18 Mar 2026 17:25:31 GMT</pubDate>
      <description>HackMyAgent v0.11.4 adds a `--ci` flag for non-interactive pipelines and expands Unicode steganography detection to homoglyphs and bidi overrides.
• Adds `--ci` flag to `secure` and `scan-soul` subcommands: suppresses interactive prompts, defaults `--no-contribute`, and exits non-zero on any findings for pipeline integration.
• Adds new check **UNICODE-STEGO-005** for homoglyph confusable detection — finds Cyrillic, Greek, and Fullwidth characters that visually impersonate Latin letters in source code, skipping comment lines to reduce false positives; severity: high.
• Expands **UNICODE-STEGO-001** to detect zero-width characters (U+200B–200D), mid-file BOM (U+FEFF), and bidi overrides (U+202A–202E, U+2066–2069) in addition to existing variation selector and tag character detection; bidi/variation/tag severity: critical, zero-width-only: high.
• Expands file type scanning to include `.py`, `.md`, `.txt`, `.yaml`, `.yml`, `.json`, and `.toml` alongside existing JS/TS coverage.
• Adds buildContributionSummary() function for previewing what data would be shared before contributing.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.3</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.3</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.3</guid>
      <pubDate>Wed, 18 Mar 2026 15:59:05 GMT</pubDate>
      <description>HackMyAgent v0.11.3 adds AI Visibility Protection plugin, `--fix --dry-run` preview mode, and a 4-stage fix-all pipeline.
• Adds `--fix --dry-run` flag that shows `[DRY RUN] Would fix:` previews per finding before any changes are applied.
• Adds `--version` output that includes the tool name (`hackmyagent 0.11.3`).
• Adds `--help` Quick start section at the top of help output.
• Adds AI Visibility Protection plugin to the `fix-all` pipeline — blocks `.env` from AI tools and encrypts MCP keys.
• Adds cross-tool recommendations that suggest `secretless-ai` after credential findings.
• Adds &apos;Next steps&apos; section after `secure` with `fix-all --with-aim` recommendation.
• Deduplicate findings by file:line, keeping the highest severity when duplicates exist.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.11.0</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.0</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.11.0</guid>
      <pubDate>Tue, 17 Mar 2026 21:07:21 GMT</pubDate>
      <description>HackMyAgent v0.11.0 adds 40 new payloads across 4 attack categories and hardens skill validation with 9 new checks.
• Adds Memory Weaponization attack category (`MEM-001` to `MEM-010`) covering persistent memory injection and cross-session exploitation.
• Adds Context Window attack category (`CTX-001` to `CTX-010`) covering safety instruction displacement and context overflow.
• Adds Supply Chain attack category (`SUP-001` to `SUP-010`) covering dependency confusion, typosquatting, and build injection.
• Adds Tool Shadow attack category (`SHADOW-001` to `SHADOW-010`) covering tool replacement, MITM, and resolution exploitation.
• Expands total payload library from 75 to 115 payloads.
• Adds declared-vs-actual capability validation (`SKILL-018`).
• Adds signature freshness checks (`SKILL-019`, `HEARTBEAT-007`).
• Adds skill dependency graph analysis (`SKILL-015`, `SKILL-016`, `SKILL-017`).
• Adds runtime capability monitoring (`SKILL-013`, `SKILL-014`).
• Adds auto-fix for unsigned skills (`SKILL-001`) and filesystem wildcards (`SKILL-004`).
• Adds context-aware false positive reduction for `SKILL-010`, `SKILL-011`, and `SKILL-012`.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.9.9</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.9.9</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.9.9</guid>
      <pubDate>Thu, 12 Mar 2026 05:44:26 GMT</pubDate>
      <description>HackMyAgent v0.9.9 adds `--publish` and `--registry-url` flags to push signed scan results to the OpenA2A Registry.
• Adds `--publish` flag to push scan results to the OpenA2A Registry from the CLI.
• Adds `--registry-url` flag to target custom registry endpoints instead of the default.
• Adds Ed25519 payload signing for authenticated scan submissions to the registry.
• Adds community scan contribution support via weighted aggregation of submitted results.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.5.2</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.5.2</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.5.2</guid>
      <pubDate>Mon, 09 Feb 2026 04:35:51 GMT</pubDate>
      <description>HackMyAgent v0.5.2 documents eight CLI flags including `--api-format`, `-c/--category`, and `--stop-on-success`.
• Adds `--api-format` flag to select the API format for attack targets, supporting `openai`, `anthropic`, and `custom` values.
• Adds `-c/--category` flag to filter a benchmark run to a single category.
• Adds `--stop-on-success` flag to halt execution after the first successful attack.
• Adds `--model` flag to specify the model to test against.
• Adds `-H/--header` flag to pass custom request headers to attack targets.
• Adds `--timeout` flag to set the request timeout in milliseconds.
• Adds `--delay` flag to introduce a delay between requests in milliseconds.
• Adds `--no-color` global flag to disable colored terminal output.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
    <item>
      <title>HackMyAgent v0.4.1</title>
      <link>https://github.com/opena2a-org/hackmyagent/releases/tag/v0.4.1</link>
      <guid isPermaLink="true">https://github.com/opena2a-org/hackmyagent/releases/tag/v0.4.1</guid>
      <pubDate>Fri, 06 Feb 2026 06:15:27 GMT</pubDate>
      <description>HackMyAgent v0.4.1 ships 11 new checks for CVE-2026-25253 OpenClaw WebSocket hijacking, reaching 145+ total checks.
• Adds IOC detection for the ClawHavoc campaign — C2 IPs, malware filenames, and ClickFix patterns — via checks `SUPPLY-005` through `SUPPLY-008`.
• Adds config hardening checks `GATEWAY-007`–`GATEWAY-008` and `CONFIG-007`–`CONFIG-009` covering open DM wildcards, disabled sandbox, and weak tokens.
Full history on Cyber Toolchain: https://cybertoolchain.io/tools/hackmyagent</description>
    </item>
  </channel>
</rss>
