The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core functionality involves ingesting and acting upon untrusted data from external CI logs.\n
Ingestion points: The skill fetches build logs and job outputs from Buildkite (scripts/fetch_buildkite_failures.py), CircleCI (scripts/fetch_circleci_failures.py), and GitHub Actions (references/github.md).\n
Boundary markers: There are no explicit delimiters or system instructions defined to prevent the agent from following instructions embedded within the logs.\n
Capability inventory: The skill encourages the agent to propose and apply code fixes based on log parsing, and can trigger the systematic-debugging skill for further action.\n
Sanitization: While the scripts use regex for pattern matching and truncate logs to 500 characters, these measures do not sanitize the content against malicious natural language instructions.\n- [COMMAND_EXECUTION]: The skill executes local git commands for environment discovery.\n
Evidence: scripts/fetch_buildkite_failures.py and scripts/fetch_circleci_failures.py use subprocess.run to call git branch --show-current and git remote get-url origin. These calls use static arguments and are restricted to retrieving project metadata.\n- [DATA_EXFILTRATION]: The skill manages sensitive API tokens and performs network operations to CI provider domains.\n
Evidence: The skill requires BUILDKITE_API_TOKEN and CIRCLECI_TOKEN to function. All identified network requests are directed to official, well-known service endpoints including api.buildkite.com, buildkite.com, and circleci.com for the purpose of fetching build artifacts.

fetch-ci-build