The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and process content from untrusted external codebases to generate documentation and audits. This creates a significant surface for indirect prompt injection, where malicious instructions hidden in the target repository could attempt to influence the agent's behavior or conclusions.
Ingestion points: The skill reads source files, documentation, manifests, and git history from the target directory.
Boundary markers: Instructions do not specify the use of delimiters or warnings for the agent to ignore embedded instructions in the ingested data.
Capability inventory: The agent can execute shell commands, run Python/Bash scripts, and write files to the system.
Sanitization: No explicit sanitization or filtering of the external code content is required before interpolation into the agent's context.
[EXTERNAL_DOWNLOADS]: The skill instructions suggest downloading and using several well-known third-party tools and packages for diagrams and security scanning.
Node.js packages: references/diagrams.md mentions npx -y @mermaid-js/mermaid-cli and npm install mermaid jsdom for diagram validation.
Security tools: scripts/stack_probe.sh checks for and runs common scanners like trivy, osv-scanner, pip-audit, and npm audit if available.
[DYNAMIC_EXECUTION]: The documentation in references/diagrams.md provides a Node.js script template (validate.mjs) for the agent or user to write and execute at runtime to perform Mermaid syntax validation.
[COMMAND_EXECUTION]: The skill utilizes local scripts (scripts/repo_census.py and scripts/stack_probe.sh) to perform analysis tasks, involving operations like file system traversal, regex searching, and calling external binaries.

reverse-engineer