The Agent Skills Directory

[SAFE]: The skill's primary operations involve reading project manifest files (e.g., package.json, Cargo.toml, pyproject.toml) and writing a generated README.md file. All operations are consistent with the stated purpose of document generation.
[COMMAND_EXECUTION]: The included validation script scripts/validate.sh uses bash -c to perform integrity checks on the SKILL.md file. The commands executed are static, local to the skill directory, and do not involve untrusted input.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from project files (Step 2) and user responses (Step 3) to populate the README template. While these files are technically untrusted external data, the skill's capabilities are limited to writing markdown content and invoking a validation council, which presents a minimal security risk.
Ingestion points: README.md, PRODUCT.md, package.json, pyproject.toml, go.mod, Cargo.toml, Makefile, LICENSE, CHANGELOG.md.
Boundary markers: None explicitly defined in the generation prompt.
Capability inventory: File write to README.md, execution of local scripts/validate.sh, and invocation of the council skill.
Sanitization: No specific sanitization or escaping of project file content is mentioned, but the output context (Markdown) limits the potential for execution-based attacks.

readme