Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
PRECC
v0.2.4-1Automatically fixes bash command errors, compresses CLI output, and redirects code search for efficient, cost-saving Claude Code execution.
⭐ 0· 88·0 current·0 all-time
byYijun Yu@yijunyu
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The description promises a single Rust binary that auto-corrects shell commands, intercepts grep/rg, and compresses agent context. The registry lists no required binaries, env vars, or config paths. That mismatch suggests the skill claims capabilities it does not declare — installing hooks and intercepting CLI tools should require explicit install and config declarations.
Instruction Scope
SKILL.md instructs the agent/user to run a remote install script (curl|bash), install a precc-hook that intercepts CLI tools, 'ingest --all' to mine session history, and compress/modify CLAUDE.md and memory files. Those runtime actions read and modify user session history and agent memory files and may redirect searches and outputs to external services — broader scope than the registry metadata exposes.
Install Mechanism
The only install instruction is piping a raw.githubusercontent.com script to bash. While GitHub raw is a known host, curl | bash executes remote code without local verification (no checksum, no signed release). There is no alternative install path, no checksums, and no local install spec in the registry — this is higher-risk.
Credentials
The skill declares no required environment variables or credentials, yet its features (semantic search, Claude/CCC/RTK integration, possible remote indexing) likely require API keys and access to agent memory files. The SKILL.md references cocoindex-code, RTK, and backups but does not declare or justify any credentials or specific config paths.
Persistence & Privilege
Although not marked always:true, the install instructions imply persistent system changes: a precc-hook that intercepts commands and automatic backups/rewrite of CLAUDE.md and memory files. That amounts to persistent system-level hooks and file modification beyond a transient skill — users should treat this as privileged and persistent installation.
What to consider before installing
This skill is internally inconsistent and requires caution. Before installing: (1) Do NOT run curl | bash blindly — inspect the install script on the GitHub repo and verify what it writes/executes. (2) Ask the author for a formal install spec, checksums/signatures for the binary, and a list of exact files/paths the tool will read or modify. (3) Verify what external services (Claude, cocoindex, RTK) require API keys and why; never provide keys without clear justification. (4) If you must test it, run it in an isolated VM or disposable container, not on a production laptop with private keys or sensitive agent memory. Providing those mitigations and clearer declarations would raise confidence; absence of them keeps this 'suspicious.'Like a lobster shell, security has layers — review code before you run it.
latestvk97fwrs1dgkk5qz2p054xsgwd1839shf
License
MIT-0
Free to use, modify, and redistribute. No attribution required.