The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common shell commands to manage version control and verify code changes via testing tools.
Evidence in prompts/branch-completion.md: Uses git checkout, git pull, git merge, git push, and gh pr create to manage branches and pull requests.
Evidence in prompts/tdd.md: Uses npm test, cargo test, pytest, and go test to execute the project's test suite as part of the TDD cycle.
[DATA_EXFILTRATION]: The workflow facilitates the transfer of local branch data to remote version control systems.
Evidence in prompts/branch-completion.md: Instructs the agent to perform git push -u origin <feature-branch> and use the GitHub CLI (gh pr create) to upload code and metadata.
[PROMPT_INJECTION]: The skill uses project-local files as a state machine to drive its automated implementation loop, which presents an attack surface for indirect prompt injection if those files contain untrusted content.
Ingestion points: The skill reads from openspec/changes/<id>/tasks.md and design.md to guide its actions.
Boundary markers: There are no explicit instructions for the agent to ignore potentially malicious directions embedded within these files.
Capability inventory: The skill has access to shell execution for git operations and testing suites.
Sanitization: No explicit verification or sanitization of the spec file content is performed before the agent processes them.

agentic-dev