The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the exec tool to perform shell-based file discovery and content inspection. Specifically, it uses ls -A to list directory contents, grep -rln to search for specific keywords within files, and cat to read the contents of existing ADR documents. These operations are scoped to the decisions/ and proposals/ directories.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a conduit for persisting user-provided narrative content into the project's filesystem, creating a surface for indirect prompt injection.
Ingestion points: User-provided descriptions of architecture decisions, technical context, and project consequences are captured and written to markdown files in SKILL.md.
Boundary markers: The skill utilizes a structured template system (via the write tool's template parameter) which provides some structural separation, though it lacks explicit instructions for the agent to ignore or sanitize embedded instructions within the user's narrative input.
Capability inventory: The skill leverages exec, write, edit, search, links, and audit tools to manage the documentation lifecycle.
Sanitization: No explicit sanitization or validation logic is defined for the content before it is processed by the write or edit tools.

record-a-decision