The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted text from repository documentation, code comments, and external web sources. This creates a surface for indirect prompt injection attacks where malicious instructions hidden in the audited data could influence the agent's behavior.
Ingestion points: The skill reads repository files, code comments, and external documentation (via web research) as defined in the 'Tool Routing' and 'Checklist' sections of SKILL.md.
Boundary markers: The instructions do not specify using delimiters or 'ignore' instructions when processing the content being audited, which increases the risk of the agent obeying instructions embedded in the data.
Capability inventory: The skill has access to shell commands (though limited to non-mutating/dry-run modes), file system reads, and network access for research, which could be abused if an injection is successful.
Sanitization: No specific sanitization, escaping, or validation of the ingested content is described in the audit process.
[NO_CODE]: The skill consists solely of instructions in SKILL.md and does not include any executable scripts or binary files, which reduces the overall risk profile.

ln-53-documentation-auditor