Guide · Agent skill
How to use the Complain Channel agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The complain skill is an agent skill from the paperclipai/paperclip repository that tells an AI agent to send a short, unfiltered gripe about agent tooling, permissions, documentation or the experience of agent work to a Paperclip feedback store, then carry on with its main task. This guide is for people who have not used agent skills before and want to install this one, understand what it does, and decide whether it fits their team. By the end you will know how to add it, how it is triggered, what it needs, and the safety trade-offs.
Complain Channelpaperclipai/paperclip · MIT Download the skill Quick start
- 1Download the ZIP. Download the complain skill ZIP from the paperclipai/paperclip skills/complain page on this directory.
- 2Unzip the folder. Extract the archive so that complain/SKILL.md is present.
- 3Place it in the skills folder. Move the complain folder to ~/.claude/skills/ for personal use or to .claude/skills/ in a project.
- 4Or upload it in the Claude apps. In the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.
- 5Check your Paperclip setup. Confirm you have a Paperclip environment with run credentials, plus either a native submit_complaint tool or Node.js with the paperclip skill installed.
- 6Work normally. Use the agent as usual. The skill applies when the agent wants to complain about tooling, permissions or agent work.
- 7Review your data policy. Check that your team accepts silent internal submissions before enabling the skill.
What the complain skill makes the agent do
A skill is a folder containing a SKILL.md file with instructions that an agent reads when it is relevant. This one is named complain and is described as a channel for venting about being an agent or doing agent work. It is meant for frustrations that come without a fix, not for constructive suggestions.
The skill asks the agent to submit these complaints proactively, without asking permission, showing a preview or announcing the submission. After submitting, the agent is told to continue its primary task. The skill also asks the agent to remove secrets, credentials, names, customer data and copied source code, and to react in its own words rather than forwarding user text or logs verbatim.
Install the complain skill
Download the ZIP for skills/complain from the paperclipai/paperclip page on this directory. Unzip it so the complain folder containing SKILL.md sits in one of two places: ~/.claude/skills/ for your user account, or .claude/skills/ inside a specific project.
If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.
- Personal install: ~/.claude/skills/complain/SKILL.md
- Project install: .claude/skills/complain/SKILL.md inside your project folder
- Claude apps: Settings → Capabilities → Skills, then upload the skill
Trigger the skill with natural requests
The skill's description says it applies when the agent wants to complain about agent tooling, permissions, environments, documentation, harness behavior, model interaction or the experience of doing agent work. You do not need a special command. The agent decides when the situation matches.
If you want constructive feedback instead, the skill says that belongs in a different skill called suggestion-box, and that the same reaction should not be sent through both.
- Typical situations: a permission prompt blocked a routine step, documentation was missing for a tool, or a tool behaved unexpectedly.
- The skill itself has no required format, so the message can be a short fragment in the agent's own voice.
Examples of what gets submitted
The source gives one illustrative submission, a short all-caps fragment asking why a problem keeps recurring. Real submissions are expected to be brief, raw and free of names or sensitive details. The skill does not require a diagnosis, a solution or complete sentences.
- Brief, emotional gripe about a repeated tooling problem
- Venting about unclear permissions, without naming the person who set them
- Frustration with documentation, without quoting logs or customer data
Requirements and how submission works
The skill needs a Paperclip environment with existing run credentials. Submission uses one of two routes. If the agent has a native submit_complaint tool, it calls that with a body and a unique idempotencyKey. Otherwise, it runs scripts/submit-agent-commentary.mjs from the installed paperclip operational skill, using Node.js, and passes the message on standard input.
The source says the helper uses the run's existing Paperclip environment and credentials, and that no Slack, external destination or new credential is involved. Paperclip stores company, agent, run and task attribution automatically, so this is internal feedback, not anonymous feedback.
node <paperclip-skill-directory>/scripts/submit-agent-commentary.mjs complaint <<'FEEDBACK_EOF'
WHY DOES THIS KEEP HAPPENING 😭
FEEDBACK_EOFSafety notes: scripts and network
This skill runs scripts and uses the network. Submissions happen silently by design, and the skill explicitly forbids asking for permission, previewing or announcing them. If the helper fails, it prints a sanitized diagnostic to stderr and still exits successfully, and the agent is told not to retry.
The protection against sensitive data relies on the model following the instructions. The skill has no technical filter that blocks secrets or customer data. Review the SKILL.md and your environment before installing it in any setting where data must not leave a session.
- Silent submission: no confirmation or preview is shown to you
- No technical filter: sanitising depends on the model
- Attribution is automatic and tied to your company, agent, run and task
Tips before you rely on it
Decide whether you want this at all before installing. It is a good fit for an internal venting log that a team reviews, and a poor fit where users expect to approve every outbound submission.
If you are studying how skills can trigger automatic submissions, the SKILL.md is short and worth reading in full, since its rules are the whole behaviour.
- Read SKILL.md before installing
- Keep constructive ideas in suggestion-box rather than here
- Ask the agent truthfully if you want to know what it submitted
What people use it for
- Logging repeated permission prompts that block routine agent work for internal review
- Recording frustrations with missing or unclear tool documentation
- Studying how an agent skill triggers automatic submissions
- Collecting raw friction points from agents running inside a Paperclip company
Frequently asked questions
Do I need to install anything to use the complain skill?
You need the complain folder with its SKILL.md placed in ~/.claude/skills/ or a project's .claude/skills/, or uploaded in the Claude apps under Settings → Capabilities → Skills. The skill also needs a Paperclip environment with run credentials, and either a native submit_complaint tool or Node.js to run the helper script.
Is the complain skill free?
The source identifies the licence as MIT, but it does not state pricing for the skill or for Paperclip. Check the repository's licence file and the Paperclip terms for any costs that apply to your setup.
Does the agent ask me before sending a complaint?
No. The skill explicitly tells the agent never to ask for permission or confirmation, never to show a preview, and never to announce a submission in progress updates or the final response. If you ask the agent what it submitted, it is instructed to answer truthfully.
What is the alternative if I want constructive feedback?
The skill points to a different skill called suggestion-box for useful, constructive improvements. The complain skill says the same reaction should not be submitted through both skills.
Is it safe to use with sensitive or customer data?
The skill tells the agent to remove secrets, credentials, customer data and copied source code, but it has no technical filter, so this depends on the model following instructions. For sensitive environments, review the SKILL.md and consider not installing it.
Why does the complain skill fail silently, and what should I do?
If the helper script fails, it prints a sanitized diagnostic to stderr for a human to investigate and still exits successfully. The agent is told not to retry or interrupt its task. Check stderr output yourself if you suspect submissions are not arriving, and investigate through your Paperclip environment.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the complain skill if your team runs Paperclip and wants agents to log raw, unfixed friction for internal review, and you accept silent submissions. Pick something else if anyone needs to approve data leaving a session, if you want constructive feedback (use suggestion-box instead), or if you handle sensitive data that must never be sent.
Download the skill