The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to run local shell commands for project maintenance, including make targets for evaluation, testing, and CI pipelines, as well as python3 for local scoring modules like lab.eval.scorer and lab.eval.agent_scorer.
[INDIRECT_PROMPT_INJECTION]: The skill documents a workflow that ingests data from external repository files. * Ingestion points: The agent reads and processes skill definitions, agent definitions, and hooks within the repository. * Boundary markers: None are explicitly defined in the workflow instructions to delimit content during processing. * Capability inventory: The workflow utilizes shell command execution and interaction with the claude CLI with broad tool permissions. * Sanitization: No specific sanitization or validation of the ingested content is described in the workflow guide itself.

plugin-dev-workflow