The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (build-chain.sh, rebase-branch.sh) and leverages a wide range of powerful CLI tools including git, gh, docker, and just to manage the development lifecycle and PR automation.
[EXTERNAL_DOWNLOADS]: The skill invokes npx playwright to capture UI screenshots. This involves the dynamic execution of a well-known external package from the npm registry at runtime.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of reading and acting upon untrusted data from the repository environment:
Ingestion points: The agent ingests data from git commit messages, file diffs, and the content of files during merge conflict resolution.
Boundary markers: There are no specified delimiters or system-level instructions to protect the agent from following malicious instructions that might be embedded within the commit history or codebase it processes.
Capability inventory: The agent possesses significant capabilities, including the ability to execute shell commands, perform network requests via curl and gh, and write to the local file system to resolve conflicts.
Sanitization: Content derived from external sources (commits and diffs) is used to generate PR descriptions and titles without evidence of sanitization or filtering to prevent the injection of malicious instructions into the agent's output or subsequent logic steps.

create-pr