show-me-your-work
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read files from the
agent-transcripts/directory to audit the log's accuracy. This introduces a surface where instructions embedded in previous conversation outputs could influence the current agent's behavior during the audit phase. - Ingestion points: Files located under the
agent-transcripts/directory. - Boundary markers: None specified for the transcript reading process; however, the output is written to a structured TSV format.
- Capability inventory: The skill executes
scripts/log.shto write to the filesystem and invokes subagents for review. - Sanitization: The
scripts/log.shscript provides robust sanitization against formula injection for all logged cells. - [COMMAND_EXECUTION]: The skill utilizes a local shell script
scripts/log.shto handle log appending. - Evidence: The script is called with six arguments to append formatted data to a TSV file. Analysis of the script shows it uses
set -euo pipefailfor safety and correctly handles untrusted input by sanitizing tabs, newlines, and potential spreadsheet formulas (prefixing with a single quote).
Audit Metadata