skills/joelhooks/joelclaw/pds/Gen Agent Trust Hub

pds

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous commands to perform administrative tasks on a Kubernetes cluster.
  • Evidence: Instructions include the use of kubectl create secret, kubectl patch svc, kubectl rollout status, and a custom joelclaw CLI for managing service health and record writing.
  • [EXTERNAL_DOWNLOADS]: The skill uses package managers to install external software and performs network requests to external APIs.
  • Evidence: Uses helm upgrade --install to deploy the nerkho/bluesky-pds release and curl to interact with the PDS management endpoints at http://localhost:9627.
  • [CREDENTIALS_UNSAFE]: The skill retrieves sensitive tokens and passwords using a secret management utility and stores them in shell environment variables for immediate use in configuration commands.
  • Evidence: Usage of secrets lease for JWT_SECRET, ADMIN_PASSWORD, PLC_ROTATION_KEY, and JOEL_PASSWORD in SKILL.md follows standard secret management practices for automated deployments.
  • [PROMPT_INJECTION]: The skill includes an interface for writing data to system logs, creating an indirect prompt injection surface if the data processed originated from untrusted external sources.
  • Evidence: The joelclaw pds write command accepts a JSON --data payload which is injected into the server records.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 05:05 PM
Security Audit — agent-trust-hub — pds