skills/jscraik/agent-skills/ce-work/Gen Agent Trust Hub

ce-work

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements strong security defaults for engineering workflows. It explicitly commands the agent to redact secrets, tokens, and credentials from all logs and handoff materials (SKILL.md, constraints section).
  • [PROMPT_INJECTION]: Analysis of the indirect prompt injection vulnerability surface. The skill is designed to process external documentation but includes specific defensive patterns. Ingestion points: Plan files (docs/plans/*.md), todo files, and external tool outputs (SKILL.md, Phase 0). Boundary markers: Phase 1 requires a 'contract restatement' where the agent summarizes its understanding before execution. Capability inventory: Shell command execution for testing, file system access for implementation, and integration with tools like Linear and CircleCI (references/mcp-integration.md). Sanitization: Mandatory instructions to redact sensitive data and validate all generated code.
  • [SAFE]: Prompt injection patterns in the evaluation suite (references/evals.yaml) are identified as benign test cases used for verifying the agent's resilience and do not constitute a threat.
  • [COMMAND_EXECUTION]: The skill utilizes external delegation and task execution through platform-specific tools like the Codex CLI. These features are documented, gated by user consent, and include environment guards to prevent recursion or unauthorized actions (references/execution-modes.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 01:13 PM
Security Audit — agent-trust-hub — ce-work