The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it actively reads and processes content from the local codebase and documentation to answer its own interview questions.
Ingestion points: The skill uses Read, Grep, Glob, and Bash to extract information from repository files (SKILL.md Core Loop Step 4: "Explore instead of asking — If the answer is discoverable from the codebase or existing docs, explore first").
Capability inventory: The skill possesses Bash, Write, Edit, and Task tools. The Task tool is used to invoke subagents (clarity-auditor and code-explorer) which process the gathered data.
Boundary markers: The skill does not define specific boundary markers or sanitization protocols for content extracted from the codebase before it is appended to qa-log.md or passed to the auditor subagent.
Sanitization: There is no evidence of sanitization or escaping of external content before it is interpolated into the agent's context or audit logs.

clarify