adobe-debug-bundle

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides a shell script (adobe-debug-bundle.sh) that automates the collection of diagnostic data. It uses npm list to check SDK versions, curl for connectivity tests, and nslookup for DNS resolution.
  • [DATA_EXFILTRATION]: The script performs an OAuth token test by sending environment variables (ADOBE_CLIENT_ID and ADOBE_CLIENT_SECRET) to the official Adobe IMS endpoint (ims-na1.adobelogin.com). This is documented as a connectivity check and targets a well-known service endpoint associated with the skill's purpose.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The script accesses the local .env file to include configuration keys in the diagnostic bundle. It uses sed to redact the values, ensuring that secrets are not exposed in the final archive.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface as it ingests data from local configuration files (.env).
  • Ingestion points: Reads from the .env file in the current directory.
  • Boundary markers: The script uses sed to redact all values, preventing instructions or secrets within the file from being processed as active content.
  • Capability inventory: Uses Bash(curl), Bash(tar), and Bash(grep) for diagnostic collection.
  • Sanitization: Employs sed 's/=.*/=***REDACTED***/' to mask all values before they are added to the summary output.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 04:44 PM