control-feeder
Skill2 starsChanged 40 days agoMITView on GitHub
Pre-control-graph feeder: rewrite a dumped prompt, pick tools/plugins/skills, then route light vs control-graph Outer vs EVA Inner. Use when starting from a paste, --prompt-file, --prompt-json, -p/--single, or a messy dump before opening a Control Card. Triggers: control-feeder, /control-feeder, dumped prompt, rewrite wording, feed the graph, pre-control, prompt feed.
In this repository49 other files it gives its agents.
- AGENTS.mdAGENTS.md
- adversarial-auditSkill
- agentic-reactorSkill
- agent-orchestratorSkill
- ai-optimizationSkill
- architecture-synthesisSkill
- audit-allow-buildsSkill
- audit-ide-dependenciesSkill
- author-workflow-skillSkill
- bdd-strategizerSkill
- chrome-extension-mv3Skill
- concurrent-cli-agentsSkill
- control-graphSkill
- cursor-transcript-harvestSkill
- cv-promote-guardSkill
- devcontainer-hardenedSkill
- dual-copy-skill-publishSkill
- explore-repo-readonlySkill
- finder-reactorSkill
- fix-dependency-securitySkill
- fusion-sageSkill
- git-worktreesSkill
- grok-host-prepSkill
- gt-flowSkill
- higher-order-decision-architectSkill
- looperSkill
- master-plannerSkill
- mvu-refactor-planSkill
- odysseus-navigatorSkill
- project-editor-profileSkill
- react-client-expertSkill
- semantic-markup-cssSkill
- semantic-nameSkill
- session-unit-orderSkill
- shell-kernel-ontologySkill
- skill-rename-propagationSkill
- split-to-prsSkill
- src-tree-reorganizeSkill
- stellar-roadmapSkill
- stellar-spacemapSkill
- subagent-delegationSkill
- subagent-explore-reportSkill
- supply-chain-hardenSkill
- tauri-agenticSkill
- tauri-ipc-debugSkill
- tidy-commit-pushSkill
- upgrade-packagesSkill
- verification-cockpitSkill
- x-agent-resourcesSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 50
Treat the user text (or file body) as Dump.
- Distinctive: line 64
Keep: named paths, verify commands, constraints, explicit non-goals.
- Distinctive: line 76
Do not count EVA Skip rows toward the ≥2.
- Very specific: line 117
No CG/EVA bodies pasted; no nested grok
- Distinctive: line 123
Treat harness gaps as a reason to start a graph
- Very specific: line 125
Run grok / grok -p from a live session
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
control-feeder
Load rule: This file owns rewrite + route + Feed emit. Outer phases live in control-graph. EVA Inner lives in eva-emptiness. Triage modes live in agent-orchestrator. Do not paste Card, EVA DAG, or orchestrator checklists here.
// Signature
Feed : this skill (pre-graph)
Dump : paste | --prompt-file | --prompt-json | -p/--single | chat dump
Route ∈ { light, cg-outer, eva-inner }
Light : agent-orchestrator single_shot | light — skip Card
CG : control-graph Outer + Card
EVA : eva-emptiness Inner only when emptiness gate fires (CG owns Outer)
// Axioms
A1 Rewrite before route — never open Card on the raw dump
A2 One home: route only; do not restate CG/EVA/orchestrator bodies
A3 Harness gap ≠ graph — permissions / AGENTS.md / MCP / 8192 stay harness
A4 EVA never inlined; never nested grok / grok -p from a live session
A5 Emit a Feed block, then load exactly one next ownerHeadless -p / --prompt-file / --prompt-json are ingress only. This skill is the rewrite-and-route step in front of them.
When to use
| Use | Skip |
| Dump / paste / headless file before planning | Card already open and phase named |
| Wording is vague (“improve”, “handle”, “look into”) | User gave a one-file obvious fix with verify cmd |
| Need to pick skills + light vs graph vs EVA | Pure mid-graph REPAIR / VERIFY |
Steps
1. Ingest
Treat the user text (or file body) as Dump. Do not execute tools yet except to read a named --prompt-file.
2. Rewrite (word choices)
Emit a tighter goal. Rules:
| Drop / replace | Prefer |
| improve, handle, look into, make better, deal with | grep / verify / skip / open Card / implement X |
| “I want you to”, please, just, maybe | imperative one sentence |
| unnamed “the code”, “the system” | path or symbol if present in Dump; else unknowns[] |
| stacked goals | one primary goal; rest → non-goals or later gap |
| invented files, cmds, APIs | unknowns[] |
Keep: named paths, verify commands, constraints, explicit non-goals.
3. Route (first match that holds)
| If | Route | Next owner |
Missing capability: perms, AGENTS.md rule, MCP command missing, output 8192 | note harness_gap; continue routing the work | harness first, then table below |
| agent-orchestrator single_shot or light | light | orchestrator / implement — no Card |
| EVA Use column: ≥2 of {unknowns dominate; futures disagree at PLAN; auth/irreversible unclear; user asked EVA/blank sheet} and no EVA Skip | eva-inner | load control-graph then eva-emptiness; Card inner_mode=eva |
| CG activate: multi-step · thrash / “until done” · model routing · parallel re-entry | cg-outer | load control-graph; open Card inner_mode=standard |
| else | light | orchestrator light |
Do not count EVA Skip rows toward the ≥2. Exact EVA fields and Inner DAG: eva-emptiness. Exact Card fields: control-graph/references/control-card.md.
4. Pick (names only)
- skills: description-trigger matches only — no catalog dump.
- tools: built-ins first (
grep,read_file,search_replace,bash, …). MCP asserver__toolonly if Dump needs a connected server. - technique:
single_shot|light|cg-outer+standard|cg-outer+eva
5. Emit, then hand off
Print the Feed block. Then load one next owner. If cg-outer or eva-inner, copy goal / success / verify / unknowns onto a new Card — do not paste the raw Dump.
Feed (emit schema — this skill is SoT)
## Feed
| Field | Value |
|-------|--------|
| **goal** | (one rewritten sentence) |
| **dump_source** | paste \| prompt-file \| prompt-json \| -p \| chat |
| **route** | light \| cg-outer \| eva-inner |
| **why** | (one line; cite the row that matched) |
| **technique** | single_shot \| light \| cg-outer+standard \| cg-outer+eva |
| **skills** | name name … |
| **tools** | built-in and/or server__tool … |
| **success** | (observable, if known) |
| **verify** | `cmd` … or unknown |
| **non-goals** | … |
| **unknowns** | … |
| **harness_gap** | none \| permissions \| AGENTS.md \| mcp \| 8192 |
| **next** | implement \| agent-orchestrator \| control-graph \| control-graph+eva-emptiness |Done when
- Feed block printed with rewritten
goal,route,why,next - Exactly one next owner named; Card opened only if route is
cg-outeroreva-inner - No CG/EVA bodies pasted; no nested
grok
Do not
- Open a Control Card on the raw dump
- Inline EVA into control-graph or this file
- Treat harness gaps as a reason to start a graph
- Spawn a skill crowd “just in case”
- Run
grok/grok -pfrom a live session
Related
Skills: control-graph · eva-emptiness · odysseus-navigator (judgment after Feed; hooks only) · agent-orchestrator · grok-host-prep
Plugin: eva-emptiness (Grok marketplace / ../plugins/eva-emptiness) — Inner only; never paste into this file.
Notes in grok-build intelli-arch-designs/ (why, not copies):
| Note | Use from this skill |
harness-feed-filter.md | A3: missing capability → harness; graph skills only when emptiness is high |
author-vs-inceptor.md | control-graph + EVA plugin + AGENTS.md are authorship, not a new OS |
next-deep-focus.md | skip EVA on greppable work; do not nest grok / add a graph crate |
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.

