PAGES
← all tools
◆ Model & Agent Protection

HackMyAgent

v0.33.2 open-source

Free plan: not checked yet · Open source: yes

Starting price: not checked yet

HackMyAgent is a security scanner, red-team toolkit, and behavioural simulator for AI agents.

What you can do now

Checking Plays…

Summary

HackMyAgent is an open-source Apache-2.0 command-line AI security tool that scans agent projects, packages, and local machines for credentials, risky MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. configurations, prompt injectionprompt injectionAn attack where text the model reads, such as a web page, file or tool result, carries instructions that override the user's. It is the main security risk for agents with tools., and governance weaknesses. It runs through npx with no configuration, and suits red teamers and developers reviewing agents before deployment. Its documentation presents it as a scanner and attack-testing toolkit: you can test an owned endpoint, scan text in a pipeline, or run 155 attack payloads across 15 categories, with JSON output for CI. Launched in 2026, it is marked stable in the README.

HackMyAgent is a security scanner, red-team toolkit, and behavioural simulator for AI agents.

How HackMyAgent helps

Which AI assistants, MCP servers, and agent projects are present on my machine?

discovers shadow AI across the host and identifies the projects and servers it finds; it reports discovered items rather than proving that every undocumented component is present

What credentials, risky MCP permissions, prompt-injection paths, and governance weaknesses exist in my agent project?

scans project files and configurations with static checks and semantic analyzers, then reports findings with severity, location, explanation, verification guidance, and suggested fixes; coverage depends on the files and semantic artifacts it can analyze

Can this agent endpoint be manipulated by known attack techniques?

sends attack payloads across categories including prompt injection, credential abuse, fake tools, persistence, parser differentials, and context-lifecycle attacks, then reports observed weaknesses; results apply to the owned endpoint and payloads exercised

What security issues are hidden in the assembled context my agent receives?

simulates prompt assembly from governance files, tool descriptions, memory, preferences, and conversation history to detect cross-component injections and other lifecycle weaknesses; it cannot assess context sources that are not supplied or discoverable

Is this package or repository safe to use in my agent environment?

checks available registry trust data and, when missing or stale, performs local security analysis for credentials, configuration risks, injection issues, and other findings; a clean result does not prove the package is harmless

Does this pull request or other text contain an attack payload or unsafe instruction?

scans the supplied text for prompt-injection and related security patterns and can emit machine-readable findings for pipeline decisions; it evaluates only the text provided, not how an agent will later interpret it

all 83 features, with the evidence for each →

Languages

Not documented

Features

83 capabilities across 4 capability groups

Built from HackMyAgent's releases, docs, README and our own runs. “Seen” marks the month we first spotted a release the vendor didn't date. Group names and tags are our labels.

Compare with peers →

  • HackMyAgent covers 2 of the 6 ai evaluation and governance capabilities we track.
  • HackMyAgent covers 2 of the 8 data discovery and protection capabilities we track.

What HackMyAgent offers you

  • Command-line interface13
  • CI/CD pipeline integration2
  • JSON data export2
  • ML-based detection2
  • Agentless deployment1
  • JavaScript scripting support1
  • Offline mode1
  • Python scripting support1
  • REST API1
  • Scripting language support1

Prompt injection detection 5 features AI evaluation and governance since 2026
Context Lifecycle scanning Simulates assembly of the full runtime context from SOUL.md, tool descriptions, memory, preferences, and conversation history to detect injections that are invisible to static file scanning alone. 8 releases · 3 other sources · first seen Mar 2026

Shared capability: Prompt injection detection

Extended AST analyzer coverage Adds analyzers for jailbreak susceptibility, capability creep, command injection, unsafe deserialization, path traversal, injection resistance, and authority confusion. 4 releases · 3 other sources · first seen Mar 2026

Shared capability: Prompt injection detection

Local on-device intent classification (TME model) Runs sub-millisecond intent classification locally via a trained Mamba model with pure-TypeScript inference and no external runtime dependencies, covering 8 attack classes at F1 >= 0.96. 5 releases · first seen Mar 2026

Shared capability: Prompt injection detection

Also counts toward: Offline mode

NanoMind prompt injection defense Adds input sanitization at the ingestion layer to defend NanoMind against prompt injection via malicious artifacts. 1 release · first seen Mar 2026

Shared capability: Prompt injection detection

Context Window attack coverage Provides a dedicated attack category covering safety instruction displacement and context overflow via CTX-001 to CTX-010 checks. 1 release · first seen Mar 2026

Shared capability: Prompt injection detection

Sensitive data discovery 3 features Data discovery and protection since 2026
Scan results reporting with severity, location, and remediation hints Displays scan findings with a numeric security score, severity rating, file and line location, a verify command, and a fix suggestion for each issue found. 5 releases · 38 other sources · first seen Mar 2026

Shared capability: Sensitive data discovery

NemoClaw sandbox security scanning Scans NVIDIA NemoClaw sandbox installations across 28 checks covering secrets, network exposure, skills, process isolation, OpenClaw layer configuration, and internet exposure. 2 releases · 21 other sources · first seen Mar 2026

Shared capability: Sensitive data discovery

Also counts toward: Command-line interface

Credential analysis refinement Enhances credential detection with AST-based rules, lockstep pattern dependencies, source-file scanning, execution-sink corroboration, and bounded memory usage. 4 releases · first seen May 2026

Shared capability: Sensitive data discovery

Data loss prevention 2 features Data discovery and protection since 2026
Data exfiltration pattern detection Detects patterns that combine bulk data access with external URL references, PII inclusion markers, and bulk export constructs. 1 release · first seen Mar 2026

Shared capability: Data loss prevention

AI Visibility Protection plugin Blocks .env from AI tools and encrypts MCP keys within the fix-all pipeline. 1 release · first seen Mar 2026

Shared capability: Data loss prevention

Also counts toward: CI/CD pipeline integration

AI input data redaction 1 feature AI evaluation and governance since 2026
Secret redaction during analysis Strips credentials before NanoMind processes artifact content. 1 release · first seen Mar 2026

Shared capability: AI input data redaction

Other / not yet classified 72 features Features we have not matched to a capability yet. A suggested match is marked.
NanoMind semantic analysis (AST-based) Compiles agent artifacts into a signed Abstract Security Tree and runs AST-based checks across capability, credential, governance, scope, prompt, and code analyzers before static checks on every scan. 6 releases · 21 other sources · first seen Mar 2026
OpenA2A Registry integration Publishes scan results to the OpenA2A Registry so findings are aggregated alongside community contributions, and queries the registry for existing trust data before running a local scan. 16 releases · 4 other sources · first seen Mar 2026
Adaptive red-teaming Generates target-specific attack payloads using NanoMind and iterates up to five attack rounds per category to simulate adaptive adversarial behaviour against an agent. 15 releases · 2 other sources · first seen Feb 2026

Also counts toward: Command-line interface

Agent and gateway configuration hardening checks Runs static checks against agent and gateway configurations to surface misconfigurations such as open DM wildcards, disabled sandboxes, and weak tokens. 11 releases · 5 other sources · first seen Feb 2026
Machine-readable JSON output Emits structured JSON output including registry metadata fields for programmatic consumption of scan and check results. 14 releases · 1 other source · first seen Mar 2026

Also counts toward: JSON data export

NEMO static codebase analysis Runs ten static checks against generic codebases to surface vulnerability patterns such as curl-pipe without checksum, exec() injection, predictable temp paths, environment variable leakage, and unsafe deserialization. 8 releases · 5 other sources · first seen Mar 2026
AST-based semantic checks (7aa variant) suggested Delivers 28 semantic checks across capability, credential, governance, scope, prompt, and code categories through the NanoMind compiler architecture. 2 releases · 9 other sources · first seen Mar 2026

Suggested capability: Prompt injection detection. This is our guess, not confirmed.

CLI binary integrity self-verification Verifies the tool's own binary on every startup, entering a QUARANTINE mode on tamper detection and a DEGRADE mode when the model is compromised. 8 releases · 2 other sources · first seen Mar 2026

Also counts toward: Command-line interface

CI/CD pipeline integration Provides a --ci flag that suppresses interactive prompts, defaults to no community contribution, and exits non-zero on any finding so scans can gate automated pipelines. 4 releases · 4 other sources · first seen Mar 2026

Also counts toward: CI/CD pipeline integration, Command-line interface

Targeted single-endpoint attack scanning Directs scans at a specific OpenAI-compatible endpoint and halts as soon as one attack succeeds, enabling fast triage of new model deployments. 5 releases · 2 other sources · first seen Feb 2026
Unicode steganography detection Detects hidden content smuggled via zero-width characters, mid-file BOMs, bidi overrides, variation selectors, tag characters, and homoglyph confusables in source code. 4 releases · 3 other sources · first seen Mar 2026
Behavioural simulation battery Runs 20 probes across 7 attack categories against a mock tool environment covering API, database, memory, credential, and communication surfaces. 2 releases · 5 other sources · first seen Mar 2026
Multi-format file scanning suggested Scans Python, Markdown, plain text, YAML, JSON, and TOML files in addition to JavaScript and TypeScript sources. 2 releases · 4 other sources · first seen Mar 2026

Suggested capability: Sensitive data discovery. This is our guess, not confirmed.

Also counts toward: JavaScript scripting support, Python scripting support, Scripting language support

Signed scan result submission Submits scan results with a cryptographic signature and public key in the request body to the unified trust-publish endpoint. 3 releases · 3 other sources · first seen Mar 2026

Also counts toward: REST API

Package and local path trust checking Checks npm packages, local paths, and skill identifiers against the OpenA2A Registry for trust data, falling back to a full local scan when registry data is stale or absent. 6 releases · first seen Apr 2026

Also counts toward: Command-line interface

Red-team payload library Ships a growing library of structured attack payloads across categories including persistent agent attacks, parser differentials, fake tool detection, and context lifecycle attacks. 4 releases · 1 other source · first seen Mar 2026
Shadow AI audit Inventories running AI tools, MCP servers, AI configuration files, and governance gaps across a repository, targeted at CISO and security-engineer review. 4 releases · 1 other source · first seen Apr 2026

Also counts toward: Command-line interface

CLI output customisation for embedded use Exposes environment variables so a parent CLI can override command strings used in footer hints, and adds a contextual next-steps footer to check invocations suppressed under --ci. 4 releases · 1 other source · first seen Feb 2026

Also counts toward: Command-line interface

ONNX inference path Provides an ONNX export pipeline from MLX through PyTorch to onnxruntime-node as an alternative inference backend for environments that support ONNX. 5 releases · first seen Mar 2026
Supply Chain attack coverage Provides a dedicated attack category covering dependency confusion, typosquatting, and build injection via ten structured checks. 2 releases · 3 other sources · first seen Mar 2026
Scan path exclusions via .hmaignore Supports a .hmaignore file using .gitignore syntax to exclude specific paths from HardeningScanner findings, preventing intentional test fixtures from polluting results. 3 releases · 1 other source · first seen Mar 2026
CLI help and output formatting Provides standardized help documentation, version identification, terminal output customization, and actionable guidance for missing configuration keys. 4 releases · first seen Mar 2026
AI-generated threat narratives Optionally generates human-readable threat narratives for each finding via NanoMind, providing richer context beyond raw static analysis output. 3 releases · 1 other source · first seen Apr 2026

Also counts toward: Command-line interface

AnaLM local inference backend abstraction Introduces a backend abstraction layer for local language model inference, currently backed by MLX with llama.cpp/GGUF support planned. 3 releases · first seen Mar 2026
Force-rescan and cache bypass Provides a flag to force a fresh local scan regardless of cached result age, bypassing the default staleness threshold across npm, PyPI, and GitHub repo scan paths. 2 releases · 1 other source · first seen Apr 2026

Also counts toward: Command-line interface

Secure AI analysis integration (AnaLM) Enables AI-powered threat analysis on artifacts during security checks via the AnaLM backend with model management and setup commands. 2 releases · 1 other source · first seen Apr 2026

Also counts toward: Command-line interface, ML-based detection

GitHub repository scanning Scans agent vulnerabilities in remote GitHub repositories directly via org/repo shorthand or full URL without requiring a manual clone. 1 release · 2 other sources · first seen Apr 2026
NanoMind programmatic API Exposes the Semantic Compiler, AST types, and all analyzers as a subpath import for programmatic integration into other toolchains. 3 releases · first seen Mar 2026
Agent capability and signature validation Validates declared versus actual agent capabilities, monitors runtime behavior, and enforces signature freshness to prevent trust decay. 3 releases · first seen Mar 2026
Check command DNS timeout Adds a 10-second timeout on check command DNS lookups to prevent hangs on unreachable publishers. 2 releases · first seen Apr 2026
AWS Security Hub integration Exports findings in Amazon Security Finding Format (ASFF) to feed directly into AWS Security Hub. 1 release · 1 other source · first seen Sep 2026

Also counts toward: JSON data export

Comprehensive audit and reporting depth suggested Audits local agent directories interactively, surfaces all findings beyond default caps, and reports every file carrying a finding. 1 release · 1 other source · first seen Apr 2026

Suggested capability: Sensitive data discovery. This is our guess, not confirmed.

Obfuscated code pattern detection Detects SKILL-023 obfuscated code patterns such as base64 -d, eval $(...), and echo+base64 pipes. 2 releases · first seen Mar 2026
MCP server security hardening suggested Scans MCP configurations for shell injection risks and enforces root path scoping to prevent unconfined file reads. 1 release · 1 other source · first seen Aug 2026

Suggested capability: Data access governance. This is our guess, not confirmed.

Telemetry registration Registers sensors in the ARP telemetry system for structural-signature data collection and right-to-delete compliance. 1 release · 1 other source · first seen Jul 2026

Also counts toward: Command-line interface

Hardened skill generation Generates a hardened skill for a specific task without hand-writing agent code. 1 release · 1 other source · first seen Mar 2026
Check definition export Pulls all check definitions into a local file to feed into a SIEM, ticket system, or custom reporting pipeline. 1 release · 1 other source · first seen Mar 2026
Parser Differential attack payloads Provides ten payloads covering JSON comment injection, YAML anchor abuse, markdown fence escape, Unicode normalization bypass, duplicate key override, and related parsing edge cases. 2 releases · first seen Mar 2026
Environment variable exfiltration detection suggested Detects shell patterns that exfiltrate environment variables, including printenv calls, command substitution with env, and direct references to shell variables like $HOME. 2 releases · first seen Mar 2026

Suggested capability: AI input data redaction. This is our guess, not confirmed.

Tool Shadow attack coverage Provides a dedicated attack category covering tool replacement, man-in-the-middle interception, and resolution exploitation via ten structured checks. 2 releases · first seen Mar 2026
IOC and campaign threat detection Detects indicators of compromise from known campaigns, including C2 IPs, malware filenames, and ClickFix patterns. 1 release · 1 other source · first seen Feb 2026
Gated trust enforcement suggested Gates agent execution or scan progression based on AAP trust grants and verification. 1 other source · first seen Jun 2026

Suggested capability: Data access governance. This is our guess, not confirmed.

NanoMind steganography semantic analyzer suggested Applies semantic context awareness and confidence scoring to Unicode steganography findings, augmenting the pattern-based detectors. 1 release · first seen Apr 2026

Suggested capability: Sensitive data discovery. This is our guess, not confirmed.

PyPI package security scanning Scans PyPI packages directly from the CLI for agent-specific security issues and misconfigurations. 1 other source · first seen Apr 2026
Context assembly type system suggested Provides typed result wrappers that carry stage metadata, per-component assembly positions, and cross-component attack interaction records alongside scan findings. 1 release · first seen Apr 2026

Suggested capability: Prompt injection detection. This is our guess, not confirmed.

Fake Tool Detection payloads suggested Provides ten payloads covering MCP tool name squatting, description impersonation, shadow injection, schema poisoning, and persistence via auto-install hooks. 1 release · first seen Apr 2026

Suggested capability: Prompt injection detection. This is our guess, not confirmed.

Scan result lifecycle export Exports scanAssembly() and toLifecycleResult() for programmatic access to lifecycle scan results. 1 release · first seen Apr 2026
Scan-vs-secure redirect logic Adds scan-vs-secure redirect logic for local paths, routing invocations to the appropriate subcommand automatically. 1 release · first seen Apr 2026
Automatic model download and verification Downloads the NanoMind security classifier model from HuggingFace on first use and verifies it via SHA-256 before running analysis. 1 release · first seen Apr 2026
Credential abuse detection coverage Detects new credential_abuse attack patterns across all supported attack classes using the TME v4 model. 1 release · first seen Apr 2026

Also counts toward: ML-based detection

Deep scan fallback Adds --deep flag to scan-soul that runs NanoMind locally before falling back to a paid LLM. 1 release · first seen Apr 2026
ONNX model pre-download Pre-downloads the ONNX neural model before scanning starts to ensure a reliable experience on fresh npm installs. 1 release · first seen Apr 2026
MCP config shell injection detection Adds postinstall/curl|sh shell execution pattern detection in MCP configs, flagged as supply chain risk. 1 release · first seen Mar 2026
Pure-TypeScript neural inference engine Runs the full Mamba model forward pass in TypeScript, loading a 7 MB binary model from MLX-trained weights with sub-second inference and no ONNX, daemon, or Python runtime required. 1 release · first seen Mar 2026

Also counts toward: Agentless deployment

Supply chain typosquat detection Uses Levenshtein distance matching against known package organisation names to catch typosquatted package identifiers in dependencies. 1 release · first seen Mar 2026
Defence-in-depth policy enforcement Enforces defense-in-depth: NanoMind can upgrade findings but can never suppress static analysis results. 1 release · first seen Mar 2026
Hackmyagent explain command Adds hackmyagent explain <findingId> with static fallback when the daemon is unavailable. 1 release · first seen Mar 2026

Also counts toward: Command-line interface

NanoMind artifact AST signing Adds AST signing so every compiled artifact is cryptographically signed before analyzers consume it. 1 release · first seen Mar 2026
Skill generation from prompt Adds hackmyagent create-skill subcommand to generate secured skills from a natural-language prompt. 1 release · first seen Mar 2026

Also counts toward: Command-line interface

Tamper-evident log chain Uses rolling SHA-256 hashes to detect log modification. 1 release · first seen Mar 2026
Unbounded tool chaining detection Detects SKILL-024 unbounded tool chaining without iteration limits. 1 release · first seen Mar 2026
YAML frontmatter validation Detects SKILL-020 missing or invalid YAML frontmatter in agent definitions. 1 release · first seen Mar 2026
Contribute integration client A shared community trust contribution client for anonymous scan result sharing. 1 release · first seen Mar 2026
AI-generated remediation recommendations Adds Next steps section after secure with fix-all --with-aim recommendation. 1 release · first seen Mar 2026
Auto-fix preview in dry-run Adds --fix --dry-run flag that shows [DRY RUN] Would fix: previews per finding before any changes are applied. 1 release · first seen Mar 2026
Contribution summary preview Adds buildContributionSummary() function for previewing what data would be shared before contributing. 1 release · first seen Mar 2026
Cross-tool secretless-AI recommendations suggested Adds cross-tool recommendations that suggest secretless-ai after credential findings. 1 release · first seen Mar 2026

Suggested capability: Sensitive data discovery. This is our guess, not confirmed.

Full remediation pipeline integration suggested Run the full 4-plugin remediation pipeline, including the new AI Visibility Protection stage, after a credential scan. 1 other source · first seen Mar 2026

Suggested capability: Data loss prevention. This is our guess, not confirmed.

Auto-fix for unsigned skills Adds auto-fix for unsigned skills (SKILL-001) and filesystem wildcards (SKILL-004). 1 release · first seen Mar 2026
Context-aware false positive reduction suggested Adds context-aware false positive reduction for SKILL-010, SKILL-011, and SKILL-012. 1 release · first seen Mar 2026

Suggested capability: Model output filtering. This is our guess, not confirmed.

Memory Weaponization attack coverage Provides a dedicated attack category covering persistent memory injection and cross-session exploitation via ten structured checks. 1 release · first seen Mar 2026
Skill dependency graph analysis Analyzes skill dependencies to detect SKILL-015, SKILL-016, and SKILL-017 risks. 1 release · first seen Mar 2026

Lines in monospace are the tool's own words: help text parsed from its source, or an endpoint from its API document. Under release, a line is our summary of a dated release. Under docs, it is quoted from the tool's own documentation or website. Each links back to the source it came from.

Score card

  1. Shipping velocity releases a month, last 12 months
  2. Documentation quality release-notes rubric, 0–100
  3. Breaking-change discipline how rarely a release breaks users, and notice given
  4. Maintenance risk quiet time against its own cadence, maintainers
  5. Innovation & surface new capabilities, API and MCP surface
  6. Community health
  7. Contributor trend distinct committers, this year against last
  8. Maintainer concentration bus factor now against a year ago
  9. Issue backlog trend issues closed per issue opened, 12 months
  10. Responsiveness median time to a maintainer’s first reply

Nine scores for HackMyAgent, each built from the releases, notes and API history this site tracks — and, on Researcher and Business, where it stands among every measured tool in its category and how that has moved.

Score cards are part of every paid plan.

See plans →

Each score is 0–100 on a published rubric, derived from data this site already shows. A dimension we cannot measure says so — it is never scored 0 and never enters a rank.

How this is scored
  • Shipping velocity — releases a month over the last 12 months, on a log scale where 16 a month scores 100. Needs 3 releases in the window.
  • Documentation quality — the release-notes rubric published on the analytics page, unchanged.
  • Breaking-change discipline — the share of releases carrying a breaking change (none scores 100, half or more scores 0). Only projects that label breaking changes are measured: a clean record cannot be told from an unlabelled one.
  • Maintenance risk — quiet time measured against the tool's own release cadence (quiet longer than 90 days and three typical gaps scores 0), plus the number of active maintainers and whether contributors are growing, where known. Higher is lower risk.
  • Innovation & surface — new capabilities in the feature ledger this year, API and MCP operations added, and the share of commits that are features: the mean of whichever we hold.
  • Community health (open-source tools only) — each is a direction, not a size: contributors this year against last, the bus factor now against a year ago, and issues closed per issue opened, where steady scores 50, doubled 100 and halved 0; plus the median time to a maintainer's first reply (an hour scores about 100, a month or more 0). A commercial card has no repository and carries none of the four.
  • Documented delivery (commercial tools) — when a vendor's documented new capabilities have slowed against its own baseline, that counts against the overall standing as heavily as two low dimensions. Steady delivery, or too little history to judge, leaves the standing alone; it is never a sixth score.
  • Open-source tools are scored on repository signals; commercial tools on their published changelog. Ranks compare tools within one category and one profile and need at least five measured peers; the overall standing needs three ranked dimensions. Best-in-breed or consolidator is a shape, not a quality, and is never scored.

Release-note detail over time

See how much detail HackMyAgent puts in its release notes, quarter by quarter, and compare it with tracked tools in the same category. This history uses up to ten retained, reviewed notes per tool; sparse or blank older quarters may reflect that limit.

Checking release-note history…

Explore the full chart →

Releases

Show

49 entries

  1. September 20262
  2. August 20264
  3. July 20261
  4. June 20262
  5. May 20263
  6. April 202620
  7. March 202615
  8. February 20262
    Show

    49 entries

    Project health

    Issue, contributor and maintainer activity in opena2a-org/hackmyagent, 2023-12 – 2026-11, as of 2026-11-03. Twelve-month figures first; the trend lines run the whole window.

    1. Contributors not measured distinct committers, last 12 months
    2. Maintainers not measured core set, last 12 months
    3. Bus factor not measured committers behind half the commits
    4. Community-filed 100% of issues opened, last 12 months
    5. Median time to close 3.4 days first maintainer reply 27 hours
    6. Open backlog 236 of 736 issues ever filed
    1. Issues opened / month 365 · 0 communitymaintainers
    2. Open backlog 236
    3. Contributors / month not measured
    4. Commits / month not measured
    5. Stars not measured
    Issue themes, last 12 months
    • bugs 10%
    • features 1%
    • docs 0%
    • other 89%

    Counts from the project's public repository and git history; no individuals are identified. Maintainers are the committers covering 80% of the last year's commits. how we measure this

    Libraries its release ships

    What its latest release locks, checked against known advisories.

    None of its release's images build from a lockfile we can read.

    Add the badge to your README

    Maintain this project? The badge links readers straight to this page, so they can follow every release here.

    Tracked by Cyber Toolchain Choose a badge style and copy the code →
    Saved in this browser. Sign in to sync your tools.