OpenClaw Agency Agents

WarnAudited by ClawScan on May 10, 2026.

Overview

This skill is review-worthy because it downloads mutable agent personas from GitHub and writes them into OpenClaw’s persistent instruction files, changing how your agent behaves.

Install only if you trust the upstream agency-agents-zh repository and are comfortable with this skill changing OpenClaw’s persistent persona/instruction files. Before activating a persona, review its markdown content, note that updates pull the latest upstream version, and keep the restore command available.

Findings (3)

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 selected or later-updated persona could steer the agent’s goals, style, tool use, or priorities beyond what the user expected.

Why it was flagged

The selected markdown file from the cloned repository is copied into AGENTS.md, which can make external persona instructions authoritative for future agent behavior.

Skill content
awk '... skip==3 { print } ...' "$FOUND_FILE" > "$WORKSPACE_DIR/AGENTS.md"
Recommendation

Review the persona markdown before activation, keep system/developer instructions higher priority than persona text, and require explicit user confirmation before replacing AGENTS.md.

What this means

If the upstream repository changes or is compromised, new instructions could be imported into the user’s agent configuration.

Why it was flagged

The skill imports content from a mutable external GitHub repository without pinning a commit, tag, checksum, or reviewed copy; update.sh also pulls from main.

Skill content
git clone https://github.com/jnMetaCode/agency-agents-zh.git "$REPO_DIR"
Recommendation

Pin the repository to a reviewed commit or release, show the exact source version to the user, and avoid automatic updates without review.

What this means

The agent may continue behaving as the activated persona until the user restores the previous configuration.

Why it was flagged

The activation process writes persistent workspace context files that can affect later sessions and tasks, not just the current command.

Skill content
cat > "$WORKSPACE_DIR/SOUL.md" ...; cat > "$WORKSPACE_DIR/IDENTITY.md" ...; ... > "$WORKSPACE_DIR/AGENTS.md"
Recommendation

Make activation state clearly visible, provide a simple restore path, and consider time-limited or session-scoped activation by default.