moai-jev-skill-suggestion
Skill1.3k starsChanged 2 months agoApache-2.0View on GitHub
Guidance for a display-only skill-suggestion capability: how a caller should suggest a skill to the operator, how the emitted answer is read, and what must never happen (selection, auto-loading, judgment). The capability gate ships off; this skill carries mechanism and contract ONLY — never the call path; where and how to reach the capability is the catalogue's concern.
- Reads credentials
Trending
More talk lately.about the whole repository
Who's talking about modu-ai/moai-adk
- Ven0m0/claude-config“Add a structured external integration plan to the root TODO backlog”GitHub · Mar 18, 2026 · 4 reactions
- Ven0m0/claude-config“TODO list”GitHub · Jan 31, 2026 · 1 reactions
- sudosubin/agents.nix“codex-marketplaces.github.modu-ai.moai-adk: 3.1.2 → 3.2.0”GitHub · Oct 23, 2026
- sudosubin/agents.nix“agent-skills.github.modu-ai.moai-adk: 3.1.2 → 3.2.0”GitHub · Oct 23, 2026
- sudosubin/agents.nix“codex-marketplaces.github.modu-ai.moai-adk: init at 3.1.2”GitHub · Oct 13, 2026
The biggest 5 of 16 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 16
Talked about on GitHub (16 times).
On GitHub · 16
- Mar 18, 2026Add a structured external integration plan to the root TODO backlog — Ven0m0/claude-config4 reactions · 4 comments
- Jan 31, 2026TODO list — Ven0m0/claude-config1 reaction · 18 comments
- Oct 23, 2026codex-marketplaces.github.modu-ai.moai-adk: 3.1.2 → 3.2.0 — sudosubin/agents.nix0 reactions
- Oct 23, 2026agent-skills.github.modu-ai.moai-adk: 3.1.2 → 3.2.0 — sudosubin/agents.nix0 reactions
- Oct 13, 2026codex-marketplaces.github.modu-ai.moai-adk: init at 3.1.2 — sudosubin/agents.nix0 reactions
11 more on GitHub
- Sep 30, 2026agent-skills.github.modu-ai.moai-adk: init at 3.1.2 — sudosubin/agents.nix0 reactions
- Sep 7, 2026Worktree-isolation Bash check folds command substitutions in quoted-delimiter heredoc bodies but refuses braces there, though such bodies cannot expand — anthropics/claude-code0 reactions · 1 comment
- Sep 6, 2026docs(t41): moai gate 종료 코드 합류 — upstream 이슈 문안 (제출 안 함) — bjw202/minidiscord0 reactions
- Jul 23, 2026Hands-on eval: promote MoAI-ADK to core tier — cameronsjo/spec-compare0 reactions
- Apr 22, 2026Add MoAI-ADK to Collections & Libraries — travisvn/awesome-claude-skills0 reactions
- Apr 22, 2026Add MoAI-ADK to Meta & Orchestration category — VoltAgent/awesome-claude-code-subagents0 reactions
- Apr 22, 2026Add MoAI-ADK — SPEC-first Agentic Development Kit — jqueryscript/awesome-claude-code0 reactions
- Apr 22, 2026Add MoAI-ADK — SPEC-first AI-assisted coding framework — wsxiaoys/awesome-ai-coding0 reactions
- Apr 22, 2026Add MoAI-ADK — SPEC-first Agentic Development Kit for Claude Code — e2b-dev/awesome-ai-agents0 reactions
- Nov 19, 2025Issue #1: Estructura base del proyecto .NET 9 — mjcuadrado/mjcuadrado-net-sdk0 reactions
- Nov 13, 2025fix: resolve conflitos Git + remove hooks SYNC bloqueantes (Windows f… — PedroGiudice/lex-vector0 reactions
In this repository66 other files it gives its agents.
- AGENTS.mdAGENTS.md
- moai-cleanSkill
- moai-codemapsSkill
- moai-e2eSkill
- moai-feedbackSkill
- moai-fixSkill
- moai-gateSkill
- moai-goalSkill
- moai-gtdSkill
- moai-harnessSkill
- moai-loopSkill
- moai-mxSkill
- moai-planSkill
- moai-projectSkill
- moai-reviewSkill
- moai-runSkill
- moai-syncSkill
- moai-todoSkill
- hns-lsel-applierSkill
- hns-lsel-curatorSkill
- hns-moaiadk-best-practicesSkill
- hns-moaiadk-dev-referenceSkill
- hns-moaiadk-patternsSkill
- hns-oss-docs-i18n-rulesSkill
- hns-oss-docs-readme-syncSkill
- hns-oss-docs-structure-mapSkill
- hns-oss-docs-verifySkill
- hns-workflow-ci-loopSkill
- moai-domain-backendSkill
- moai-domain-databaseSkill
- moai-domain-design-dnaSkill
- moai-domain-frontendSkill
- moai-domain-html-reportSkill
- moai-domain-humanizeSkill
- moai-domain-svg-infographicSkill
- moai-factory-foremanSkill
- moai-foundation-ccSkill
- moai-foundation-coreSkill
- moai-foundation-qualitySkill
- moai-foundation-thinkingSkill
- moai-harness-learnerSkill
- moai-lane-watchdogSkill
- moai-meta-harnessSkill
- moai-ref-api-patternsSkill
- moai-ref-aside-browserSkill
- moai-ref-cross-model-auditSkill
- moai-ref-git-workflowSkill
- moai-ref-jev-question-designSkill
- moai-ref-llm-securitySkill
- moai-ref-owasp-checklistSkill
- moai-ref-react-patternsSkill
- moai-ref-secopsSkill
- moai-ref-seoSkill
- moai-ref-supply-chainSkill
- moai-ref-testing-pyramidSkill
- moai-ref-ui-polishSkill
- moaiSkill
- moai-workflow-dddSkill
- moai-workflow-docs-claim-checkSkill
- moai-workflow-loopSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 51
Not a selection. The existing selector keeps its selection authority.
- Distinctive: line 53
Not an auto-load. No skill content is loaded, attached, or injected because a suggestion was emitted.
- Distinctive: line 55
Not a judgment. Not a completion decision, not a merge approval, not a queue mutation — anything hard to undo stays with people.
- Distinctive: line 57
Not a code-path input. Nothing branches on the answer.
- Distinctive: line 79
Every item below is guidance conditioned on a caller existing — it ships no behavior:
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
Skill Suggestion, Display-Only
The gate comes first
The capability is gated behind workflow.jev.enabled, and the shipped
default is off. Confirm the gate before anything else. With the gate off, a
caller constructs no request, makes no network call, and proceeds with no
signal: an absent signal is NO SIGNAL — the same as "we did not ask" — never
a negative answer, and never a reason to change course.
Three readiness surfaces exist for the on state: the gate key itself, the
credentials file referenced by convention (~/.moai/.env.typesafe, read by
the credential loader — never inlined, never copied into settings), and the
doctor readiness check (a TCP-only probe that reports gate, credential, and
reachability in three lines without sending any judgment request). A caller
that cannot confirm all three proceeds as if the gate were off.
The reachable surface itself lives in the product catalogue, not here. This skill owns mechanism and contract only; the catalogue owns where the call lives.
What a suggestion is
A suggestion is a signal a person reads: a labelled model answer that names a skill that might be worth loading and carries a confidence, and it does nothing else. Four things it is not:
- Not a selection. The existing selector keeps its selection authority; a suggestion never flips it.
- Not an auto-load. No skill content is loaded, attached, or injected because a suggestion was emitted.
- Not a judgment. Not a completion decision, not a merge approval, not a queue mutation — anything hard to undo stays with people.
- Not a code-path input. Nothing branches on the answer; a human reads it, and the reading is the only consumer.
The loading mechanisms rejected here
Two mechanisms were evaluated and rejected; recording them keeps them from returning quietly:
- Hiding the listing and injecting a chosen SKILL.md body directly. This breaks the display-only chain: a skill that never appeared in context has no reading human, and a direct content injection is precisely the auto-load the contract forbids.
- An ambient function-hook wiring that evaluates a suggestion on every prompt. It commercializes a model call per prompt on an unmeasured behavior, fragments the template distribution, and rides an early-experiment surface that can shift without notice. Not adopted. Revisit only with measured evidence that a constant-answer baseline has been beaten — the measurement bar belongs to a separate track, not to this skill.
Operational checklist (for when a caller exists)
Every item below is guidance conditioned on a caller existing — it ships no behavior:
- Log each decision, and announce once per session that suggestion logging is active — one visible line per session, not a chorus per call.
- Filter non-task origins. A prompt typed by the operator is not a task event; only genuine task origins may produce a suggestion.
- Filter invocation-disabled skills. A skill marked disable-model-invocation stays invisible to suggestions; the filter runs before the call, never after the answer.
- Cache the roster and skill bodies per session; re-reading listings and bodies on every prompt is the waste a session cache exists to prevent.
- Guard against duplicate injection: a skill already present in context is never suggested to appear again. Check what is loaded before showing a name.
- Canonicalize names: a display name and a directory id are two spellings of one skill; map both through one canonical form before deduplicating, logging, or displaying.
Question design is a separate discipline
The suggestion call is a judgment over a typed question, and typed questions carry their own rules — compute the question in code, always admit a no-match answer, keep the supplied state small, phrase both polarities. Those rules live in the question-design reference skill; this skill does not restate them. The measurement regime above the surface is likewise owned elsewhere: this skill records the contract, never baseline numbers.
What this skill is not
This skill is not a capability. It ships no caller, registers no command, adds no hook, and changes no setting. It is not a verdict: suggestion answers never close cards, approve merges, or mutate queues. And it is not an availability promise: the gate ships off, and nothing in this skill makes a suggestion surface exist — the checklist above governs only if a caller is ever built elsewhere, and until then an absent suggestion is just NO SIGNAL.
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.

