mine
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script (
scripts/digest.py) usingpython3 -I. This script further invokes theopencodeCLI tool to interact with session data. These executions use argument lists rather than shell strings, preventing typical command injection vulnerabilities. - [INDIRECT_PROMPT_INJECTION]: The skill processes historical session logs as input, which represents a surface for indirect prompt injection. The skill implements specific mitigations:
- Ingestion points: Data is fetched from the OpenCode API and written to temporary markdown files.
- Boundary markers: Instructions in
SKILL.mdexplicitly direct the agent to treat processed digests as 'evidence, not instructions'. - Capability inventory: Capabilities are limited to file reads, local temp file writes, and reporting findings to the user.
- Sanitization: The
digest.pyscript uses a comprehensive regular expression (SECRET) to identify and redact sensitive credentials (API keys, tokens, passwords) before the logs are processed by the LLM.
Audit Metadata