signal-friction-scan

Pass

Audited by Gen Agent Trust Hub on Aug 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the vendor-owned joelclaw CLI along with standard utilities like hostname, find, and rg. These tools are used to query local session metadata and search for patterns within the user's repository paths (specifically ~/Code/joelhooks/).- [PROMPT_INJECTION]: The skill ingests and processes agent session transcripts, which represent a surface for indirect prompt injection from potentially untrusted interaction data.
  • Ingestion points: Session content is retrieved through joelclaw sessions signals, joelclaw sessions friction, and joelclaw session inspect in SKILL.md.
  • Boundary markers: The instructions explicitly advise treating session signals as "radar, not truth" and emphasize synthesis over literal execution of instructions found within history.
  • Capability inventory: The skill possesses the capability to execute shell commands and write files to the vendor's repository structure during its second pass.
  • Sanitization: The process uses a two-pass workflow where the first pass is strictly observational (synthesis only), and the second pass (creating or patching assets) requires an explicit request from the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 9, 2026, 05:04 PM
Security Audit — agent-trust-hub — signal-friction-scan