The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, creating a surface for potential indirect prompt injection.
Ingestion points: The instructions in SKILL.md (Step 1) require the agent to read the PR's title, body, linked issues, and reviews.
Boundary markers: There are no explicit instructions to use delimiters or ignore instructions embedded within the PR metadata.
Capability inventory: The skill has the capability to perform repository writes (git rebase, git merge), network operations (git push), and post comments.
Sanitization: The skill includes a 'secret/private-data check' before posting authorized comments, providing some mitigation against data exposure.
[COMMAND_EXECUTION]: The skill performs powerful version control operations that can alter repository history.
Evidence: Instructions in SKILL.md (Step 4) specify the use of git push --force-with-lease for remote updates.
Safety Controls: The operating contract requires the agent to stop for real blockers, specifically before history rewrites or changing data outside the repository, unless explicitly authorized by the user.

modernize