The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the repository (both documentation and source code) to determine which facts are redundant.
Ingestion points: The workflow in SKILL.md describes scanning the repository for instruction files (e.g., AGENTS.md, .cursorrules) and performing broad searches across source and configuration files.
Boundary markers: The prompt template for the reviewer agent in SKILL.md includes explicit instructions to ignore the specific documentation files being audited to ensure a blind test.
Capability inventory: The skill in SKILL.md spawns sub-agents, performs repository searches using grep and glob, and possesses file-write permissions to rewrite documentation files.
Sanitization: The instructions do not specify explicit sanitization or validation of the content discovered during code searches before it is used to score and potentially remove documentation facts.
[COMMAND_EXECUTION]: The workflow involves spawning sub-agents that use search tools such as grep and glob and read system files to verify fact discoverability. These operations are core to the skill's functionality and are restricted to the codebase context.

agents-md-lint