- Home
- Skills
- Prompt & Skill Engineering
- Task Observer: One Skill to Rule Them All
Task Observer: One Skill to Rule Them All
The meta-skill that builds and improves all your skills, including itself.
Free
Works with the AI Agents you already use
Task Observer: One Skill to Rule Them All
Example session with this skill installed
Any observations logged this session?
- Read your context and instructions
- Compiled the task observer
5 observations logged this session. All are logged for your next scheduled review; nothing has been changed in any skill file.
Improvements to existing skills
| Skill | What happened | Proposed change | Type |
|---|---|---|---|
| conversation-defaults | Twice this session, something was reported as missing on the strength of a tool that summarises rather than shows: a spreadsheet connector that rendered 11 of 13 rows without saying so, then a page fetch that paraphrased. You corrected both. | Second violation of the same rule, so no rewording: an absence claim must name the raw artefact it was checked against, or it does not ship. | open-source |
| outreach-pipeline | The "is the brand already listed?" check ran against the repository a website names as its source. The site's list and the repository's list turned out to be different lists. | Run the check on the page that was actually cited, never on its claimed upstream. | open-source |
| outreach-pipeline | A temporary import file was left for you to delete, as a line in the closing summary. | Log temporary files as they are created and delete them as a step of the run, not a line of its report. | open-source |
| api-companion | A tool's required argument is named differently from the field its own discovery call returns, so the first call failed. | Document the argument name, not only the concept. | open-source |
| outreach-pipeline | A cap written to limit incremental prompt tuning got in the way of a deliberate scope expansion. | State the rule's purpose, so structural changes are recognised as outside it. | internal |
Principle captured
A tool that answers a question about an artefact cannot establish a negative about it. Presence can be established by a summary; absence cannot.
Next steps
- Your next scheduled review picks these up automatically.
- To act on one now instead, say "act on #1626" and a staged update is prepared for you to review.
Connects securely to your tools. The creator never sees your data.
About this skill
Creating skills is powerful but time-consuming, and the skills that do get built stay frozen: they never learn from how you actually use them.
Task Observer is a meta-skill that runs alongside your work, watches what you do, and does two things:
-
It identifies new skills for you. It spots repeating patterns in your work, flags them as skill candidates and helps you build them, so you get skills without staring at a blank page.
-
It improves your existing skills. It notices corrections you make, preferences you express and gaps in your current skills, then suggests specific updates.
It also watches itself. If it misses something, triggers where it shouldn't, or its own observation format could be clearer, it logs that too. The skill that improves all your skills also improves itself.
The observer never modifies your skills directly. It produces recommendations that you review, so you stay in control of what changes and when. Patterns that apply across skills are captured as cross-cutting principles, and new or updated skills are checked against them.
Who it's for: anyone who uses skills and wants them to get better over time instead of staying frozen. You don't need to be a developer. One honest boundary: its value compounds with scale. With a handful of skills, your assistant's built-in memory may cover much of the same ground; once editing skills directly stops feeling manageable, the observation log and review cycle pay off.
Track record: over 1,500 observations logged across 82 skills, most of them turned into skill improvements. The current version includes improvements suggested by 65 different users across 141 issues and 33 pull requests.
Open source under CC BY 4.0: github.com/rebelytics/one-skill-to-rule-them-all
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skill Creator Skills.
Built for Claude: other agents need some adaptation, because the content assumes Claude's architecture. Activation needs one extra step: description matching alone under-triggers, so add the activation line to your CLAUDE.md (or your platform's equivalent) and check in a new session that it runs. The full experience needs filesystem access (Claude Cowork, or Claude Code in the desktop app). The formal log and review cycle pay off most as your skill library grows; with a handful of skills, your assistant's built-in memory may be enough.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
11 uses
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
- Software
Node.js
Python
Permissions
File Scopes
Shell: runs short bash snippets and one helper script to derive observation ids and create log files. Read and write files: only its own observation log and staged skill updates in your workspace folder. It never modifies your installed skills directly, and it never needs to fetch an external URL.