The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a shell pipeline in SKILL.md using find, sort, grep, and sed to list Markdown files and their headings within the docs/product and docs/architecture directories. This command is restricted to local relative paths and is used for project indexing.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by instructing the agent to read the full content of Markdown files from the repository. While this is the intended functionality, it creates an attack surface for instructions embedded within those files to influence agent behavior.
Ingestion points: Files located in docs/product/ and docs/architecture/ as specified in SKILL.md.
Boundary markers: None identified; instructions direct the agent to "read each in full" without explicit delimiters or warnings to ignore embedded commands.
Capability inventory: The skill possesses shell execution capabilities (the file listing script) and the ability to read arbitrary project files.
Sanitization: No sanitization, validation, or escaping of the document content is performed before the agent processes it.

consult-descriptions