The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted data from external files (other SKILL.md files) to perform scoring and inference.
Ingestion points: Processes all SKILL.md and legacy .md files found in .claude/skills/**, ~/.claude/skills/**, and .claude/commands/**.
Boundary markers: No explicit instructions are provided to the agent to treat the content of these files as data rather than instructions, nor are delimiters suggested.
Capability inventory: The skill uses Read, Glob, and Bash (specifically for find operations) to process these files.
Sanitization: There is no mention of sanitizing or escaping the content of the audited skills before they are processed for the final report, creating a surface where a malicious skill could influence the agent's output during the audit.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform file discovery using the find command. While these commands are hardcoded in the execution instructions, the use of Bash combined with user-provided directory arguments presents a standard surface for command parameter manipulation if not handled carefully by the agent.

eval-skills