The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill includes instructions to retrieve Kubernetes secrets and decode them from base64 into the terminal output, which exposes sensitive credentials to the agent's session and potential log storage.
Evidence: kubectl get secret fireflies-secrets -o jsonpath='{.data.api-key}' | base64 -d in SKILL.md.
[COMMAND_EXECUTION]: The skill executes high-privilege management commands on the production infrastructure, including modifying secrets and environment variables, and restarting services.
Evidence: kubectl create secret generic fireflies-secrets --from-literal=api-key=NEW_KEY and kubectl set env deployment/fireflies-integration RATE_LIMIT_MODE=queue in SKILL.md.
[EXTERNAL_DOWNLOADS]: Fetches service status information from status.fireflies.com and health metrics from internal application endpoints.
[PROMPT_INJECTION]: The skill processes untrusted external data from logs and API responses, creating an attack surface for indirect prompt injection.
Ingestion points: Reads error logs via kubectl logs and fetches network data via curl in SKILL.md.
Boundary markers: None present to distinguish between instructions and data.
Capability inventory: Permission to run kubectl, curl, and Bash commands.
Sanitization: No sanitization or validation of the retrieved logs or API content is performed.

fireflies-incident-runbook