The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown and YAML data from the project's prompts directory. This creates a surface where instructions embedded within prompt fragments could theoretically influence the agent's behavior during the audit. The skill mitigates this by providing findings to the user for manual review and approval rather than executing automated changes.
[COMMAND_EXECUTION]: The skill uses local CLI tools such as promptbook, wc, diff, and rg to perform structural and semantic analysis of the codebase. These commands are typical for development workflows and are scoped to the project's local files.

promptbook-doctor