The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read requirements from GitHub issues and comments to build a rule promotion manifest. While this involves ingesting external, untrusted data, the skill includes several safety measures and the ingestion is central to its legitimate purpose.
Ingestion points: GitHub issue bodies and comments (SKILL.md).
Boundary markers: The skill does not define specific prompt delimiters for external content but instructs the agent to 'Verify every listed current rule exists' on the local filesystem.
Capability inventory: The skill uses just (a command runner) to execute local repository scripts, performs file moves, and updates source code via generation tools (SKILL.md).
Sanitization: The agent is explicitly told to 'Resolve incomplete or malformed requirements before editing' and to avoid inferring missing information, which mitigates the risk of executing unintended promotion plans.
[COMMAND_EXECUTION]: The skill utilizes several just commands (just promote-rules, just gen-rules, just f, etc.). These are project-specific build and maintenance tasks intended for use by developers within the Biome repository (github.com/biomejs/biome) and do not represent a security risk in this context.

promote-lint-rules