The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads an authentication token from a local configuration file (~/.config/teamcity-access-token) to interact with the vendor's internal TeamCity API. This is handled with explicit security warnings to the user to ensure the token is stored securely and not disclosed within the chat session.
[COMMAND_EXECUTION]: The skill performs shell commands using git and gh to manage repository metadata and automate the creation of fix branches and PRs. These operations are transparent and necessary for the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from GitHub PR metadata and TeamCity test failure logs, which represents a potential surface for indirect prompt injection.
Ingestion points: PR metadata via gh pr view and test failure details via curl to the TeamCity REST API.
Boundary markers: None explicitly defined in the prompt instructions when delegating to the sub-agent.
Capability inventory: Bash (git, gh, curl) and the Agent tool for sub-agent delegation.
Sanitization: The skill uses jq to filter, projection, and truncate the JSON responses from TeamCity, ensuring only relevant failure information is processed.

fix-e2e-tests