The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using python3 -m doclifecycle. These commands incorporate parameters derived from external inputs, such as record digests and user names, which are selected from audit reports.
[PROMPT_INJECTION]: The skill processes data from external files (report.json, plan.json) and incorporates text from these files into documentation. This creates a surface for indirect prompt injection where malicious instructions embedded in audit reports could attempt to influence the agent.
Ingestion points: Audit reports, record digests, and edit plans provided as external JSON artifacts.
Boundary markers: The skill contains explicit instructions to treat replacement text as "byte-verbatim" and enforces a rigid sequence of validation steps to prevent the agent from interpreting data as new instructions.
Capability inventory: The skill can execute shell commands via the doclifecycle module and perform write operations on the local file system within the repository scope.
Sanitization: Relies on the external doclifecycle engine to validate the integrity and authority of approval sets and plans before execution.

fixing-docs