The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/sonar.py uses subprocess.run to execute git rev-parse HEAD. This is used to resolve the current repository SHA for recording in the sonar-scan-summary.jsonl attestation file. The command is static and does not incorporate untrusted input.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted issue messages fetched from the Sonar REST API, creating an attack surface for indirect prompt injection.
Ingestion points: Untrusted issue messages are retrieved from the /api/issues/search endpoint in scripts/sonar.py and scripts/sonar_rest.py.
Boundary markers: The skill explicitly implements a containment strategy in scripts/sonar.py, where the untrusted Sonar message is quarantined under the raw_input.{message} field of the finding store. This prevents the agent from processing the message as instructions during the fetch phase.
Capability inventory: The skill can perform network operations via a REST client to the configured Sonar instance and execute git via subprocess.
Sanitization: Untrusted data is promoted to top-level fields only after a separate validation and ingestion pass (manage-findings ingest), ensuring separation between data retrieval and triage processing.

workflow-integration-sonar