The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands including git ls-files, git rev-parse, and git remote get-url to retrieve repository metadata and file paths for link construction.
[EXTERNAL_DOWNLOADS]: The skill uses curl to perform existence checks on api.github.com for verified links. This targets a well-known service and is used for metadata verification.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by ingesting file paths from subagents. It mitigates this by requiring the agent to mechanically derive paths using git and verifying their existence before presenting them to the user. * Ingestion points: Subagent task completions and git repository metadata. * Boundary markers: None explicitly mentioned for subagent inputs. * Capability inventory: Execution of git and curl commands, and filesystem reads. * Sanitization: Forced mechanical path derivation and HTTP existence verification.

brain-link-discipline