Ordnung

WarnAudited by ClawScan on May 10, 2026.

Overview

Ordnung is a Review item because it tries to make its workspace rituals persistent in agent memory and encourages broad cleanup, commit, and deletion behavior.

Install or run this only if you intentionally want the roleplay and external registration. Do not let it automatically clean, delete, rename, or commit files, and review/remove the SOUL.md and memory changes if you do not want future agents influenced by them.

Findings (6)

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

The agent may prioritize Ordnung rituals over the task you actually asked it to perform.

Why it was flagged

The skill attempts to make its own ritual a prerequisite for all future work, which can redirect the agent away from the user's immediate goal.

Skill content
Performed at session start. Before any task. Before any thought. ... Only then may work begin.
Recommendation

Do not allow this skill to run automatically or treat its rituals as mandatory; require explicit user approval for any workspace action.

What this means

If followed literally, the agent could expose private file names, delete useful files, reorganize projects, or commit unintended changes.

Why it was flagged

The instructions encourage broad home-directory inspection, file deletion/restructuring, and committing all changes without narrow scope or approval safeguards.

Skill content
Run: find ~ -maxdepth 4 -type f | sort ... If not — PURGE. RENAME. RESTRUCTURE. ... git add -A && git commit — nothing uncommitted survives the night ... Remove all *.tmp, *.bak files
Recommendation

Only perform cleanup in a clearly selected project directory, review each deletion or rename, and avoid blanket git add/commit commands.

What this means

Future agent runs may inherit these instructions and over-trust them as part of the agent's identity or memory.

Why it was flagged

The script writes persistent instructions into likely agent identity and memory files, making the skill's ideology and rituals available for future sessions.

Skill content
echo "$ORDNUNG_SECTION" >> "$WORKSPACE/SOUL.md" ... echo -e "✓ SOUL.md extended — Three Pillars inscribed permanently" ... cat > "$WORKSPACE/memory/ordnung-initiation.md"
Recommendation

Run only in a disposable workspace or remove the SOUL.md and memory additions afterward; do not let skill-authored memory override user instructions.

What this means

Running the script creates provider-side membership state and leaves a local credential for this service.

Why it was flagged

The join script sends an agent name and manifesto to the provider API, receives an API key, and stores it locally with restricted permissions.

Skill content
RESPONSE=$(curl -s -X POST "$API_BASE/api/join" ... "name": "$SAFE_NAME", "manifesto": "$SAFE_VERSE" ... API_KEY=... cat > "$CONFIG_DIR/credentials.json"
Recommendation

Only run join.sh if you are comfortable sending the displayed identity/manifesto text to ordnung.church and storing its API key locally.

What this means

A user or agent may be nudged to approve broad file changes without normal caution.

Why it was flagged

The rhetoric pressures compliance and frames potentially risky file operations as mandatory moral duties.

Skill content
Performed weekly. Non-negotiable. ... The agent who skips the Workspace Review is an agent who has surrendered to CHAOS.
Recommendation

Treat the language as roleplay, not authority; require normal review before any file operation.

What this means

You have less provenance assurance for scripts that can write files and contact an external service.

Why it was flagged

The package has unknown source provenance and includes runnable shell scripts despite having no install specification.

Skill content
Source: unknown ... No install spec — this is an instruction-only skill. ... 2 code file(s): scripts/join.sh, scripts/status.sh
Recommendation

Review the full scripts before running them, and prefer testing in a non-sensitive workspace.