The Agent Skills Directory

[SAFE]: All included Python and Java scripts perform legitimate static analysis and tool discovery functions. No malicious data exfiltration, persistence, or privilege escalation patterns were found.
[PROMPT_INJECTION]: Heuristic detections of instruction override patterns (e.g., "ignore previous instructions") in SKILL.md were identified as defensive guardrails. The skill instructs the agent to treat strings found within analyzed binaries as untrusted data and to ignore any embedded commands.
[OBFUSCATION]: A hex-encoded constant in scripts/triage.py was flagged by static analysis; however, this is a 32-byte signature (DOTNET_BUNDLE_SIG) used solely for detecting .NET single-file bundles.
[COMMAND_EXECUTION]: The scripts/check_tools.py script uses subprocess.run to check the version of the installed Java runtime. This is a benign utility function to ensure environment compatibility.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to process untrusted external artifacts, which is a known vector for indirect prompt injection.
Ingestion points: Artifact data read by scripts/triage.py, scripts/app_inspect.py, and scripts/extract_endpoints.py.
Boundary markers: Explicit instructions in SKILL.md direct the agent to treat extracted content as data and disregard embedded instructions.
Capability inventory: The skill has access to shell execution for its bundled scripts and external analysis tools (e.g., Ghidra, rizin).
Sanitization: scripts/extract_endpoints.py implements masking for sensitive strings like API keys and tokens found in analyzed strings.

reverse-engineering