The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted project files to determine license compliance, creating a surface where malicious file content could influence agent output.
Ingestion points: Project manifest files (package.json, requirements.txt, pyproject.toml, go.mod) and dependency metadata in node_modules are read and parsed by the bin/audit.sh helper script.
Boundary markers: The skill does not implement explicit boundary delimiters or "ignore embedded instructions" warnings when processing the content of these external files.
Capability inventory: The skill executes local scripts, runs shell utilities (jq, awk, grep), and calls an external stack utility to save artifacts.
Sanitization: The skill normalizes input using tr and uses jq for extraction, which provides technical sanitization against command injection in the script, but the final aggregated results are passed back to the AI agent for interpretation.
[DYNAMIC_EXECUTION]: The skill locates and executes scripts using paths resolved at runtime from environment variables.
Evidence: The SKILL.md file uses the SKILL_DIR and NANOSTACK_ROOT environment variables to execute bin/audit.sh and the external utility bin/save-artifact.sh.
[COMMAND_EXECUTION]: The skill executes several shell commands and standard utilities to scan different development environments.
Evidence: The bin/audit.sh script uses jq, awk, grep, sed, and tr to identify dependencies and classify licenses for Node.js, Python, and Go projects.

license-audit