stelow-workflow-execution-critique

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local command-line utilities including git, sem, and cymbal to perform code diffing, structural analysis, and dependency mapping. These tools are used to establish a source of truth for the audit process.
  • [DYNAMIC_EXECUTION]: The skill uses inline Node.js and Python 3 scripts to process structural data and update local project files. Specifically, it generates a Node.js script to update stelow.json by incorporating findings from the execution critique, effectively automating project management state changes based on agent-generated findings.
  • [INDIRECT_PROMPT_INJECTION]: The skill includes a 'Context Audit' mode that utilizes a browser tool to ingest content from live URLs. This creates a surface where external, untrusted content could include instructions aimed at manipulating the audit report or decision matrix.
  • Ingestion points: Data retrieved from external websites via agent_browser (documented in SKILL.md).
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to treat browser-retrieved content as data rather than instructions.
  • Capability inventory: The agent has permissions to write to the file system, execute shell commands, and run dynamic scripts (node, python3).
  • Sanitization: The instructions do not specify any validation or sanitization steps for content retrieved from external URLs before it is processed for the audit.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 09:05 PM
Security Audit — agent-trust-hub — stelow-workflow-execution-critique