The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository being analyzed (including source code, documentation, and configuration manifests like package.json, pyproject.toml, and Cargo.toml) to generate documentation. A malicious repository could contain instructions designed to influence the agent's behavior.
Ingestion points: The skill reads project source files, READMEs, and build configurations to discover stack context and patterns (SKILL.md, Step 4; read_only_commands.md, Dependency Discovery).
Boundary markers: The skill uses a management marker  to identify and protect specific sections in the AGENTS.md file, providing clear boundaries between agent-managed and user-managed content (SKILL.md, Step 5).
Capability inventory: The skill uses bundled Python scripts that invoke shell commands (git, tokei) via subprocess.run and specifically instructs the agent to identify and execute type-check commands found within the codebase (e.g., tsc, mypy, cargo check) to verify changes (Working Agreements, Code Change Rules).
Sanitization: The skill lacks explicit instructions for the agent to sanitize or validate the content of discovered type-check commands before execution, relying on the agent's internal safety guardrails.
[COMMAND_EXECUTION]: The skill performs automated repository analysis by executing shell commands through bundled Python scripts. These scripts use secure methods for command invocation, such as passing arguments as a list to subprocess.run to prevent shell injection vulnerabilities.
scripts/loc_to_limit.py executes tokei to measure the repository size for character limit calculation.
scripts/git_ownership_signals.py executes git commands to identify high-churn files and ownership boundaries.
scripts/detect_monorepo.py uses localized file inspection and regex to identify monorepo structures.

agents-md-generator