The Agent Skills Directory

[COMMAND_EXECUTION]: The skill leverages the gh (GitHub CLI), awk, and mktemp utilities to automate the creation, modification, and validation of pull requests. These tools are used to implement word count constraints and ensure PR bodies are correctly formatted. The skill demonstrates defensive programming by recommending the use of quoted heredocs (<<'EOF') when passing potentially large or untrusted text blocks to shell commands, which prevents accidental shell expansion or command injection during the validation process.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow includes scanning the target repository for existing pull request and issue templates in various locations (e.g., .github/, docs/). It also involves reading existing PR bodies to verify content before offering them for human review. This ingestion of untrusted external content represents a surface for indirect prompt injection.
Ingestion points: Template detection logic in SKILL.md and PR body retrieval via gh pr view in the 'Re-read Before the Human Reads' section.
Boundary markers: The skill requires the agent to explicitly ask the user for permission before using templates found in a repository, providing a human-in-the-loop validation step.
Capability inventory: Access to the file system (via mktemp and shell redirects) and the network (via gh CLI for interacting with GitHub repositories).
Sanitization: The skill mandates a strict redaction process defined in references/redaction.md, which requires the agent to strip PII, credentials, and machine-specific paths from all output before publishing.

pr-and-issue-discipline