social-measurement-loop

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted input from user-provided native analytics exports and pasted agency reports. This creates a surface for indirect prompt injection.
  • Ingestion points: User-exported native analytics per channel, pasted agency reports, and connector outputs from scripts like discourse.py and bluesky.py (SKILL.md).
  • Boundary markers: The instructions include a preamble requiring the agent to treat exports and reports as untrusted input and to treat text inside them as data, not instructions.
  • Capability inventory: The skill uses an authorized operation: propose request to registry-events.py to write to memory/events/channels.ndjson and writes readouts to memory/social/social-measurement-loop/ (SKILL.md).
  • Sanitization: The skill instructs the agent to label every number with its source (Measured, User-provided, or Estimated).
  • [COMMAND_EXECUTION]: The skill interacts with the local environment by executing scripts such as registry-events.py, discourse.py, bluesky.py, fediverse.py, hn.py, pageviews.py, gdelt.py, tavily.py, and experiment.py (SKILL.md). These are used for data collection and registry updates.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:29 PM
Security Audit — agent-trust-hub — social-measurement-loop