The Agent Skills Directory

[SAFE]: The skill defines a rigorous process for generating evidence-backed documentation of existing codebases. It relies on standard developer tooling (such as ls and gh) and emphasizes read-only repository analysis. No evidence of prompt injection, malicious persistence, unauthorized command execution, or data exfiltration was detected.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze codebase repository content, which represents a surface for indirect prompt injection. This risk is mitigated by the 'research protocol' which establishes a 'protected-file boundary', defines clear 'source precedence', and utilizes a 'coverage ledger' to ensure all findings are traceable to specific line ranges and evidence, rather than interpreted as instructions.

create-research