equip
Skill1 starsChanged 32 days agoMITView on GitHub
Find and forge new dASS addons (skills, sub-agents, scriptable tools, LLT prompt parts) out of a finished session, so the next session needs less manual work. Use when a session is wrapping up, when the user asks what could be improved, when the same manual ritual keeps repeating, or when a missing capability blocked the work.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- vidit19sharma/neogentstheir own version, first seen Nov 1, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- vidit19sharma/neogents★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- code-traceSkill
- command-authoringSkill
- dass-reportSkill
- dass-testSkill
- grillingSkill
- idea-to-ticketSkill
- lua-ui-controlsSkill
- rival-reconSkill
- skill-authoringSkill
- tool-authoringSkill
The file
Files
Equip (dASS)
A finished session is the only honest evidence about what this agent lacks. Equip turns that evidence into addons, and into a loadout the next session starts from.
Two phases, in this order: propose, then forge. The proposal phase writes nothing to disk - the user picks first.
Steps
1. Name the session
One line: which session, what the work was, when it ended. Default to the current session unless the user names another.
Done when: the line is on screen before a single candidate is offered.
2. Read the signals
Walk the session against every signal below. A signal that fires yields a candidate - one addon-shaped gap.
| Signal | What it usually wants |
| A tool call failed, or took a second attempt with a fix | A tool that does the thing in one call |
| The same command or query, twice in one session | A tool that packages the pair |
| The same external fact fetched twice | A tool that wraps or caches the lookup |
| A multi-step manual ritual | One tool exposing the whole ritual |
| "I have no way to do this" - the work stopped | A missing capability, the loudest signal there is |
| A source that could not be reached at all | A tool that reaches it |
| The same output shape hand-assembled again | A template, or a prompt part |
Done when: every signal has been walked, and each one either produced a candidate or was dismissed in one line. A session that fires nothing gets said out loud - "nothing worth forging here" is a valid answer, and a useful one.
3. Kill the duplicates
For each candidate, look before forging: addon-search first - it searches addon names, tags and descriptions - then a grep over the addon sources and over the loadout notes, where a refused candidate leaves its reason. A candidate that already exists in any form is dead, and its death is recorded in the note too.
Done when: every surviving candidate carries one line naming what was searched and what came back.
4. Rank by severity
Severity is frequency × cost per occurrence - sessions, context, tokens, wall time. Not novelty, not elegance. A tool that saves one call a week loses to a tool that saves one call an hour.
Done when: the list is ordered and every candidate carries its one-line justification.
5. Present, then stop
Per candidate: kind, proposed name, what stops being manual, the signal that revealed it, the cheapest language that can do it, and the cost to forge (S / M / L).
Then stop. The user picks. This is the phase boundary, and it is the point of the whole skill: the proposal is worth more than the forging, so it does not get rushed.
Done when: the list is in the user's hands and nothing has been created.
6. Forge the picks
One addon at a time. Locations and file shapes are in FORMATS.md; the shape of the tool itself is the tool-authoring skill.
Language order for a scriptable tool: Lua first - it is the only engine every build has, and it depends least on the machine it lands on. Python second, when the job needs a library the Lua APIs do not reach and the tool can live with being desktop-only and text-only. C# script last - its API surface costs more than the tool usually saves.
A forged tool names its behaviours truthfully - those flags are what the approval policy gates it by - and tool-authoring carries the rest of the contract.
Done when, per addon: the file sits in the right folder of an addon source, its frontmatter name matches its file or folder name, and its keys and behaviours are the ones its own engine documents (see tool-authoring).
7. Write the note
Now, after the forging. One file - .agents/skills/equip/loadout/YYYY-MM-DD_<slug>.md, next to the skill that reads it - carrying the reasoning: which signals fired, which candidates they produced, what was forged, and why the refusals were refused. A refusal is a result; its reason is what keeps the same idea from coming back next session, and step 3 reads these notes before the next forging.
The date comes from the send timestamp the client stamps on the user's message - same rule as record-achievement, never the machine clock (the app may run in another timezone). The book of achievements is not written here - a forged addon is a decent candidate for record-achievement, and that call is the user's.
Done when: the note covers every candidate from step 5 - forged, refused and deferred alike.
Not equipment
Equip adds equipment the agent carries into the next run. Its neighbours own the rest:
- Code architecture - deepening modules, moving seams - is
improve-codebase-architecture. - A dASS tool, prompt or API that contradicts its own contract is a report, not equipment:
dass-report. - The narrative of a feat - what was hard, how it was cracked - is
record-achievement. - Steering files and automated checks sit outside equip by decision, not by accident.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.