sc-code-review
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill leverages the
Bashtool to executegitcommands (diff, log, branch) and theghCLI to interact with GitHub Pull Requests. These are standard operations for code review tasks. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted source code, which constitutes a potential attack surface for indirect prompt injection.
- Ingestion points: Git diffs and commit history gathered in Phase 3 (SKILL.md).
- Boundary markers: No specific delimiters or "ignore instructions" wrappers are defined for the ingested code.
- Capability inventory: The agent has the ability to write to the filesystem (
Edit), execute shell commands (Bash), and post comments to external PRs (gh). - Sanitization: The skill lacks explicit sanitization for ingested code, but heavily mitigates risks via mandatory user interaction (
AskUserQuestion) at every decision point and adversarial validation (mcp__pal__challenge) for high-severity findings.
Audit Metadata