skills/qianchengjie/skills/sliced-dev/Gen Agent Trust Hub

sliced-dev

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests untrusted data from the repository (code, diffs, and logs) during task preparation and review.
  • Ingestion points: Reads plan.md, decisions.md, audits.md, and project files during the READ_CONTEXT and AI_REVIEW stages.
  • Boundary markers: Explicit defensive instructions in EXECUTION-RULES.md and REVIEWER-SUBAGENT.md require agents to treat processed content as data and watch for injection.
  • Capability inventory: File modification, shell command execution, and subagent spawning.
  • Sanitization: Uses structured JSON and task briefs to limit subagent context.
  • [COMMAND_EXECUTION]: The skill executes shell commands like git, lint, and test via the dev-plan.mjs script.
  • [NO_CODE]: The logic for referenced scripts like dev-plan.mjs was not provided, precluding an audit of the underlying execution logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 10:26 AM
Security Audit — agent-trust-hub — sliced-dev