The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the local environment and external sources, creating a surface for indirect instructions to influence the agent's behavior.
Ingestion points: Reads git diffs, untracked files, commit messages, and external wiki or tracker pages (SKILL.md Phase 1).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined for processing these inputs.
Capability inventory: The skill has the ability to execute shell commands (git, linters, build tools) and write to the local file system to apply fixes.
Sanitization: No sanitization or escaping of ingested data is mentioned before it is processed by the agent.
Mitigation: The skill includes a "No silent fixes" hard rule and requires the user to approve every proposed plan and modification (Hard Rule 3, Phase 3), which serves as a critical human-in-the-loop checkpoint.
[COMMAND_EXECUTION]: The skill executes various shell commands to gather context and perform automated code quality checks.
Evidence: Executes git diff, git status, git log, and git blame to analyze the codebase (SKILL.md Step 1 & 4).
Evidence: Dynamically discovers and runs project-specific build commands, type-checkers, and linters defined in files like package.json, Cargo.toml, or pyproject.toml (SKILL.md Step 3).
Evidence: The skill explicitly verifies that tools are installed before attempting to run them, reducing the risk of unexpected execution errors.

review-and-fix