The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a wide range of potentially untrusted data from project documentation, memory banks, and knowledge trees (e.g., docs/*.md, .memory/wiki/, .agents/context/). This creates an attack surface where malicious instructions could be embedded in the audited files.
Ingestion points: Markdown documentation, memory stores, and workspace files identified in Step 1 (Scope Discovery).
Boundary markers: The procedure uses a structured 7-step approach, but there are no explicit XML or unique delimiters used to isolate processed file content from the agent's control logic.
Capability inventory: The skill has access to bash, write_to_file, replace_file_content, grep, and glob across all scripts.
Sanitization: There is no evidence of sanitization or escaping of external content before interpolation into reports or remediation steps.
[COMMAND_EXECUTION]: The skill utilizes the bash tool to perform asset extraction and integrity scans (Step 2 and 3). While these tools are standard for a repository auditing skill, they represent a high-privilege execution environment when combined with the ingestion of untrusted file content.

audit