The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project manifest files which could contain malicious instructions for the agent.\n
Ingestion points: The bin/audit.sh script reads package.json (dependency names and license types), requirements.txt, pyproject.toml, and go.mod.\n
Boundary markers: No explicit delimiters or instructions to ignore embedded content are provided in the skill instructions.\n
Capability inventory: The skill executes shell scripts, reads and writes local project files, and updates the .nanostack/config.json configuration file.\n
Sanitization: While audit.sh uses jq to safely build JSON output, the content of the manifest files is not sanitized to prevent downstream prompt injection by the AI agent.\n- [DYNAMIC_EXECUTION]: The skill executes scripts from paths computed at runtime using environment variables, which is a standard pattern for the framework it belongs to.\n
Evidence: SKILL.md defines NANOSTACK_ROOT and SKILL_DIR environment variables with defaults and subsequently executes scripts such as "$NANOSTACK_ROOT/bin/resolve.sh" and "$NANOSTACK_ROOT/bin/save-artifact.sh".\n- [COMMAND_EXECUTION]: The skill relies on shell scripts and command-line utilities to perform its license auditing tasks.\n
Evidence: The skill invokes bin/audit.sh and bin/smoke.sh, and utilizes system tools like jq, grep, awk, and sed across its operation phases.

audit-licenses