The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the GitHub CLI (gh) and Git to automate branch management. This includes high-privilege actions such as merging Pull Requests using the --admin flag. The skill includes strong safety documentation warning against using this flag to bypass failed CI, mitigating the risk of shipping broken code.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external sources, specifically Slack bot messages and GitHub PR metadata (titles and descriptions). A malicious actor could attempt to embed instructions within a PR to influence the agent's backporting logic.
Ingestion points: Slack channel C09K9TPU2G7 and GitHub PR details retrieved via gh pr view (SKILL.md, reference/discovery.md).
Boundary markers: No explicit boundary markers or safety delimiters are used when processing the text content of PRs or Slack messages.
Capability inventory: The skill possesses significant capabilities, including shell execution (git, gh, pnpm) and Python execution for conflict resolution (SKILL.md, reference/execution.md).
Sanitization: The skill lacks sanitization for the external data it processes, relying instead on its internal categorization rules to skip certain types of PRs.
[COMMAND_EXECUTION]: The skill uses Python regex snippets to programmatically resolve git merge conflicts. The logic is hardcoded and used specifically for standard conflict markers (<<<<<<< HEAD), which is a legitimate automation technique for minor code conflicts.

backport-management