show-me-your-work
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to audit and review session transcripts to verify the accuracy of the decision log. This process involves processing potentially untrusted data that may have been ingested into the session from external sources.
- Ingestion points: The agent is instructed to read the session record and transcript through the harness interface or specified file paths (
SKILL.md,references/show-me-your-work-procedure.md). - Boundary markers: The instructions specify using the active harness interface and provider schemas, but lack explicit delimiters for transcript content to prevent accidental instruction following.
- Capability inventory: The skill has the capability to write to local files (
decisions.tsv) and spawn subagents for secondary verification. - Sanitization: The logging script (
scripts/log.sh) strips control characters and prefixes formula-triggering characters with a single quote to mitigate downstream injection risks in spreadsheets. - [COMMAND_EXECUTION]: The skill utilizes a local shell script,
scripts/log.sh, to append well-formed rows to a decision log. This script is a functional part of the skill's distribution and implements input sanitization to protect against CSV/spreadsheet injection attacks when the logs are reviewed by humans.
Audit Metadata