The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub PRs, such as titles, bodies, and comments.
Ingestion points: scripts/pr-status.sh and scripts/merge-gate.sh fetch PR metadata and comments using the gh CLI.
Boundary markers: Output is typically processed via jq and rendered for human or agent consumption without explicit delimiters.
Capability inventory: The skill can execute Git operations (commit, push, merge) and GitHub API calls via gh.
Sanitization: Structured parsing with jq is used, but raw text from PR fields is rendered in status reports.
[COMMAND_EXECUTION]: Several scripts execute shell commands to interact with Git and the GitHub CLI.
scripts/conflict-marker-gate.py uses subprocess.run() to execute git diff and git show. The commands are static and used for local repository verification.
scripts/pr-status.sh and scripts/pr-merge.sh use gh api and git commands extensively. These are standard operations for the skill's intended purpose.
[EXTERNAL_DOWNLOADS]: The documentation references several well-known tools and services.
References to downloading difftastic from GitHub releases and installing frameworks like lefthook or pre-commit from official registries.
Integration with sonarcloud.io for quality gate checks is documented.

git-workflow