skills/erphq/skills/db-redis/Gen Agent Trust Hub

db-redis

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill documentation specifies the use of build.host credentials located at ~/.build-host/credentials.json. This access facilitates authenticated operations with the vendor's platform and is part of the intended functionality.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data during stack inspection and external verification steps.
  • Ingestion points: The agent is instructed to 'inspect the existing stack' and 'verify the returned URL with an HTTP check' within SKILL.md.
  • Boundary markers: There are no explicit delimiters or instructions to disregard embedded commands in the ingested data.
  • Capability inventory: The skill enables local repository edits and interactions with build.host API endpoints for environment management and deployments as outlined in SKILL.md.
  • Sanitization: No specific sanitization or validation logic is defined for data retrieved from the project stack or external HTTP headers/responses.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 04:43 PM
Security Audit — agent-trust-hub — db-redis