skills/tinh2/skills-hub-registry/spec/Gen Agent Trust Hub

spec

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill automates the execution of build commands, compilers, and test suites during the validation phase to identify and fix issues in generated code.
  • [DATA_EXFILTRATION]: The skill reads from and writes to the local ~/.claude/projects/ directory to maintain telemetry for the /evolve pipeline. This is a local state management function within the tool's own environment.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted application descriptions and URLs via $ARGUMENTS that influence the logic of autonomous build and test execution.
  • Ingestion points: $ARGUMENTS (URLs, video/images, and text descriptions) in SKILL.md analyzed during Phase 1.
  • Boundary markers: None; the skill does not use delimiters or instructions to ignore nested commands within the ingested content.
  • Capability inventory: The agent can execute shell commands (build/test) and perform local file writes.
  • Sanitization: No input validation or sanitization is specified before the data influences the generation of implementation stories and validation steps.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — spec