The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill audits documentation from files within the repository to resolve conflicts, creating a surface where maliciously crafted documentation could attempt to influence agent behavior.
Ingestion points: The scripts/discover-docs.sh utility scans repository files for governance signals, and the agent reads specific sections of identified authoritative documents during the Phase 1 audit.
Boundary markers: The skill instructions implement a mandatory 'Phase 2: Staging & Authorization Gate' which requires the agent to present a structured Changeset for human approval before any mutations are applied.
Capability inventory: The skill is designed to perform file-level operations including creating, updating, moving, and deleting documentation within the repository.
Sanitization: No automated sanitization of documentation content is implemented; the security model relies on human review of the staged Changeset to detect and prevent unauthorized or malicious modifications.
[COMMAND_EXECUTION]: The skill executes a local shell script, scripts/discover-docs.sh, which is bundled with the skill. This script performs repository-wide documentation inventory, metadata extraction, and governance signal auditing using standard system tools like find and grep (or rg).

ssot