The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell-level commands to verify the state of the local filesystem and repository integrity.
Evidence: In the 'retry loop' section of SKILL.md, the agent is instructed to "run git status in the brain repo" to check for uncommitted pages.
Evidence: The skill explicitly uses internal system tools such as gbrain get <slug> and gbrain write-request <request_id> to validate content and commit status.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes report data which could originate from external or untrusted sources, creating a surface for indirect instructions to influence the agent.
Ingestion points: The skill loads markdown files from the reports/ directory and processes results from the search tool to perform quality checks.
Boundary markers: There are no specified boundary markers or delimiters (such as XML tags or clear 'ignore' instructions) to separate report content from the agent's internal reasoning.
Capability inventory: The skill possesses capabilities to write files (put_page), execute subprocesses (git status), and query information (search, get_page).
Sanitization: No sanitization or validation of the ingested report text is performed before it is analyzed by the agent during the 'Actionability Gate' phase.

reports