percolator-bounty

Fail

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill accesses highly sensitive private key material by reading from the default Solana CLI keypair path at ~/.config/solana/id.json.
  • Evidence: Found in src/cli.ts within the loadKeypair function which attempts to read and parse the JSON secret key from the home directory.
  • [COMMAND_EXECUTION]: The skill performs unauthorized file system writes outside of its own directory by creating folders and copying files to the user's home directory (~/.openclawd/goals).
  • Evidence: The register-goal command in src/cli.ts uses mkdirSync and copyFileSync to persist a goal document to the user's personal configuration space.
  • [COMMAND_EXECUTION]: The skill implements a path traversal pattern when locating source documents for registration.
  • Evidence: In src/cli.ts, the code attempts to access a path three levels above the execution directory (../../../goals/percolator-bounty.md), which may target files outside the intended skill sandbox.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 9, 2026, 10:45 PM
Security Audit — agent-trust-hub — percolator-bounty