supply-chain-attack-recon
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: Mentions historical remote code execution patterns, such as the Codecov bash uploader compromise, as educational references. It also provides commands to install security tools like zizmor via cargo.\n- [EXTERNAL_DOWNLOADS]: Performs metadata queries against well-known registries including npm, PyPI, RubyGems, and Go proxy to verify package availability.\n- [COMMAND_EXECUTION]: Utilizes a wide range of CLI tools including gh, curl, jq, grep, and docker to automate the collection of reconnaissance data from public repositories and container registries.\n- [DATA_EXFILTRATION]: Provides instructional examples of how an attacker might exfiltrate sensitive data, such as environment variables or GitHub tokens, using curl commands within compromised workflows.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface due to its ingestion of untrusted data from external sources.\n
- Ingestion points: GitHub API responses for package manifests and workflow definitions (Step 3, 6); public JavaScript bundles (Step 3).\n
- Boundary markers: Not implemented in the provided logic.\n
- Capability inventory: Includes curl, gh, jq, docker, and secret-scanning tools.\n
- Sanitization: Primarily relies on jq and grep for data extraction and filtering.
Audit Metadata