The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize content from an entire codebase, including manifests, configuration files, and source code snippets, which are untrusted data sources.
Ingestion points: Codebase entry points, manifests, READMEs, and source code files as specified in Step 1 of SKILL.md.
Boundary markers: The skill references external documentation (shared/references/grounded-evidence.md) for grounding but does not define internal delimiters to isolate instructions from the codebase content during processing.
Capability inventory: The agent can write files to the local directory (docs/) and execute a Python validation script.
Sanitization: The skill relies on external conventions and the validate_explain_architecture.py script, which checks for external links (https?://) and placeholders to ensure the output is self-contained and safe.
[COMMAND_EXECUTION]: The skill requires the execution of a Python script to validate the output HTML file.
Evidence: SKILL.md Step 4 instructs the agent to run python3 "$SKILL_DIR/scripts/validate_explain_architecture.py" <output-file.html>. While the script performs benign regex checks, it involves executing a subprocess that takes an agent-generated file path as an argument.

explain-architecture