The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the agent to use the --admin flag when merging pull requests via the GitHub CLI (gh pr merge --admin). This allows the agent to bypass repository branch protection rules, such as required status checks or reviews.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external pull requests (titles and numbers) using gh pr list. This information is used to populate the body of new pull requests and influence the consolidation logic. A malicious actor could potentially craft a pull request title designed to influence the agent's output or behavior during the consolidation process.
Ingestion points: External data is ingested from gh pr list output in SKILL.md.
Boundary markers: None identified; PR titles are used directly in documentation and command flows.
Capability inventory: The skill has the capability to execute shell commands, manage Git branches, and perform GitHub API operations.
Sanitization: There is no explicit sanitization or validation of the ingested PR titles before they are used in the prompt for the consolidated PR description.
[COMMAND_EXECUTION]: The skill performs several potentially destructive repository operations, including resetting branches (git checkout -b, gh api ... --method PATCH -F force=true) and merging code. These actions are performed on the hardcoded repository mckinsey/agents-at-scale-ark.

ark-dependabot-management