evidence-upload

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local shell scripts (evidence-init.sh, evidence-upload.sh) and the aws CLI tool to facilitate file uploads and repository setup. These commands are integral to the skill's utility in a development workflow.- [DATA_EXFILTRATION]: Visual evidence, such as screenshots and GIFs, is uploaded to external platforms including GitHub and S3-compatible storage. These network operations are intended for the skill's primary function of evidence hosting.- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests and processes data from local results and external browser sessions without explicit sanitization or boundary markers.\n
  • Ingestion points: Local files results.json, .flowchad/config.yml, and the GitHub web interface via Navvi browser automation.\n
  • Boundary markers: Instructions do not define specific delimiters or instructions to ignore instructions embedded within the ingested files or browser content.\n
  • Capability inventory: Shell script execution, AWS CLI operations, and automated browser interactions capable of performing account-level actions.\n
  • Sanitization: There is no documented validation or sanitization of file paths, repository identifiers, or content retrieved from the browser.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 07:47 PM
Security Audit — agent-trust-hub — evidence-upload