Catl Harness Pr

AdvisoryAudited by Static analysis on May 9, 2026.

Overview

No suspicious patterns detected.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A GitHub token could be exposed to the agent transcript or logs and then used to access or modify repositories.

Why it was flagged

A literal grep can print GitHub token values from the .env file into terminal output, logs, or agent context, even though the surrounding guidance says not to echo tokens.

Skill content
先查是否已存了 GitHub 凭证(`grep GITHUB_ ~/.openclaw/.env`);但**不要自己往 .env 里写新 token**
Recommendation

Change this to a presence-only or redacted check, such as `grep -q '^GITHUB_' ~/.openclaw/.env` or a command that prints only variable names, not values.

What this means

If that PAT is leaked or misused, it could affect many repositories, not just the CATL Harness fork.

Why it was flagged

For a single CATL Harness fork/PR workflow, recommending a token with read/write access to all repositories is broader than necessary.

Skill content
Repository access:选 `All repositories` 最省事... Permissions: `Contents: Read and write` ... `Pull requests: Read and write`
Recommendation

Use the least-privileged option: create the fork first, then issue a fine-grained PAT limited to that fork where possible, and avoid broad classic `repo` or all-repository write scopes.

What this means

A misconfigured or unexpected remote could receive private repository contents during sync.

Why it was flagged

When an existing local repo has an unknown `origin`, the script only warns and later may push `main` to that same origin instead of stopping for confirmation.

Skill content
warn "origin 指向未知 URL: $current_origin" ... warn "如果这是预期,跳过;否则手动修复。" ... run "git push origin main"
Recommendation

Stop on unknown remotes unless the user explicitly approves, and validate expected HTTPS and SSH fork URL forms before any push.

What this means

If those values contain shell metacharacters, the script could execute unintended local commands.

Why it was flagged

The helper uses `eval` with command strings containing environment- or argument-derived values such as repo paths and GitHub usernames.

Skill content
run() { ... eval "$@" ... } ... run "git clone $FORK_URL $REPO_DIR"
Recommendation

Remove `eval`, use shell arrays with properly quoted arguments, and validate environment-provided paths/usernames before running Git commands.