The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes npx to download and run the @lousy-agents/cli tool from the npm registry for file discovery. This tool is a resource provided by the skill's author and is central to the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes and modifies agent instruction files, which serves as a potential vector for indirect prompt injection if the files being processed contain malicious instructions targeted at the revising agent.
Ingestion points: Repository instruction files such as AGENTS.md, CLAUDE.md, and harness-specific rule files found in .claude/rules/ or .github/instructions/.
Boundary markers: No explicit delimiters are specified to separate the instructions being revised from the agent's own processing environment.
Capability inventory: The skill uses Bash (for git and vendor tooling), Read, Glob, and Grep to manage and modify files.
Sanitization: Input prose from instruction files is processed without specific sanitization or filtering logic.
[COMMAND_EXECUTION]: The skill executes git commands (git diff, git log, git show) to evaluate changes in instruction prose and history. It also uses rm to clean up temporary metadata files generated during discovery.

instruction-style