The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted sources within the local codebase and incorporates it into persistent instruction files, creating a surface for indirect prompt injection.
Ingestion points: The skill scans project files including README.md, package.json, source components, and brand documentation to infer design patterns.
Boundary markers: There are no explicit instructions or delimiters defined to prevent the agent from obeying malicious instructions that might be embedded within the analyzed project files.
Capability inventory: The skill utilizes file-reading capabilities to scan the codebase and file-writing capabilities to update .impeccable.md and .github/copilot-instructions.md.
Sanitization: The skill lacks a sanitization or validation step to filter out malicious content or instruction overrides from the files it reads before persisting them into the agent's configuration.

teach-impeccable