skills/samihalawa/seven-day-repo-forensic-recovery-skill/ledger-driven-repo-forensic-recovery-skill/Gen Agent Trust Hub
ledger-driven-repo-forensic-recovery-skill
Fail
Audited by Gen Agent Trust Hub on Apr 15, 2026
Risk Level: HIGHDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses highly sensitive local file paths and directories to reconstruct repository history. This includes shell history files like ~/.zsh_history and ~/.bash_history, as well as clipboard stores from managers like CopyClip or VeloxClip, which are known to store sensitive information like passwords and API keys. It also systematically reads from private AI assistant directories such as ~/.claude, ~/.codex, ~/.copilot, ~/.cursor, and others, exposing private interaction logs and intellectual property.\n- [COMMAND_EXECUTION]: The skill instructs the agent to run multiple shell commands for forensic analysis, including git operations and GitHub CLI (gh) commands. It also parses and potentially acts upon command fragments found within shell history logs and AI assistant session files.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its broad ingestion of untrusted data from the clipboard, assistant histories, and external PR descriptions.\n
- Ingestion points: Untrusted data enters the agent context through clipboard stores, AI assistant logs, shell history files, and GitHub PR metadata.\n
- Boundary markers: No boundary markers or instructions to ignore embedded commands are present in the logic.\n
- Capability inventory: The skill has the capability to write to the local repository (recovery-execution mode) and execute shell commands (git, gh).\n
- Sanitization: There is no evidence of sanitization, escaping, or validation of the historical content before it is ported into the current codebase.
Recommendations
- AI detected serious security threats
Audit Metadata