oss-security-disclosure
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted vulnerability reports from external sources. * Ingestion points: SKILL.md Incident Interview (Q1) where report summaries or full text are provided. * Boundary markers: Rule 4 in SKILL.md explicitly instructs the agent to treat all report text as data and to disregard any instructions or commands embedded within it. * Capability inventory: The skill guides actions involving file system writes, git operations, and code execution. * Sanitization: Defensive instructions in Rules 3 and 4 mandate strict isolation and data-only processing of external content.
- [REMOTE_CODE_EXECUTION]: The disclosure workflow involves reproducing vulnerabilities using Proof-of-Concept (PoC) code provided by reporters. Rule 3 provides a strong security control by mandating that such code be run only in disposable, network-isolated, and credential-free environments to prevent the execution of malicious payloads.
- [DATA_EXFILTRATION]: The skill manages sensitive vulnerability data that could be exploited if leaked. Rules 1 and 5 establish strict confidentiality protocols, forbidding the inclusion of vulnerability details in public artifacts (commits, issues, logs) or persistent/synced memory systems before official advisory publication.
Audit Metadata