The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill functions by analyzing untrusted external content such as source code repositories, architectural descriptions, and legal documents, which establishes a surface for indirect prompt injection attacks.
Ingestion points: The agent reads and extracts data from these artifacts in step 2 to build a privacy inventory, as detailed in SKILL.md and references/code-review-checklist.md.
Boundary markers: The skill does not provide specific instructions to use boundary markers or delimiters when processing untrusted content.
Capability inventory: The skill utilizes filesystem search tools and writes a final compliance report to the user's local directory.
Sanitization: The instructions in SKILL.md explicitly require the agent to exclude real values of personal information from the report, focusing instead on categories and locations to minimize data exposure.
[COMMAND_EXECUTION]: The references/code-review-checklist.md file provides specific ripgrep (rg) command patterns for the agent to use to locate sensitive identifiers such as Social Insurance Numbers (SIN) and health card information within a codebase.
[SAFE]: The skill references several official and well-known government resources, including the Office of the Privacy Commissioner of Canada (priv.gc.ca), Justice Laws (laws-lois.justice.gc.ca), and the Commission d'accès à l'information du Québec (cai.gouv.qc.ca), for its compliance mapping and guidance.

pipeda-law25-compliance