system-forge
Skill10 starsChanged 2 months agoMITView on GitHub
Design a multi-capability system as a swarm, vertical, or plugin; compile it only when a Foundry runtime is available. Use when several skills or agents need explicit topology, shared state, conflict resolution, termination, deployment adapters, and cross-surface packaging.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- frankxai/Starlight-Intelligence-Systemword for word, first seen Sep 6, 2026
Other versions1 project uses a version of this.
Word for word · 1
- frankxai/Starlight-Intelligence-System★ 10100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
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 repository49 other files it gives its agents.
- AGENTS.mdAGENTS.md
- site/AGENTS.mdAGENTS.md
- .claude/hooks/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- site/CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/editorial.mdcCursor rule
- .cursor/rules/starlight-agents.mdcCursor rule
- .cursor/rules/starlight-core.mdcCursor rule
- .cursor/rules/starlight-memory.mdcCursor rule
- .cursor/rules/starlight-voice.mdcCursor rule
- frank-brand-editorSkill
- agent-hygieneSkill
- estate-guardSkill
- SKILL.mdSkill
- crypto-intelligenceSkill
- agent-forgeSkill
- skill-forgeSkill
- taste-engineSkill
- integration/ecosystem-syncSkill
- integration/repo-bridgeSkill
- integration/transmission-protocolSkill
- integration/universal-adapterSkill
- intelligence/decision-frameworkSkill
- intelligence/hermes-searchSkill
- intelligence/pattern-recognitionSkill
- intelligence/strategic-reasoningSkill
- intelligence/systems-thinkingSkill
- heartSkill
- storageSkill
- memory/capture-disciplineSkill
- memory/context-preservationSkill
- memory/insight-distillationSkill
- memory/knowledge-synthesisSkill
- memory/memory-consolidationSkill
- sis-memory-orchestratorSkill
- memory/vault-managementSkill
- agent-creatorSkill
- orchestration/context-engineeringSkill
- orchestration/gencreator-stackSkill
- orchestration/hermes-swarmSkill
- orchestration/multi-agent-coordinationSkill
- orchestration/parallel-executionSkill
- orchestration/workflow-designSkill
- workflow-skill-creatorSkill
- safety/mutation-approvalSkill
- safety/permission-gateSkill
- safety/private-public-splitSkill
- safety/secret-detectorSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 8 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Do not create a swarm merely to make a workflow sound advanced.
- Distinctive: line 23
Start with a Task Envelope.
- Distinctive: line 26
Prefer registered skills and agents.
- Distinctive: line 27
Treat lexical graph suggestions as discovery hints, never silent routing decisions.
- Distinctive: line 39
Write success conditions, stop conditions, maximum rounds, cost or latency ceilings, and escalation points.
- Distinctive: line 40
Keep builder, critic, and verifier separate when stakes or ambiguity justify it.
- Very specific: line 42
Foundry runtime available: only when tools/foundry/cli.mjs and foundry/contracts/ exist in the current workspace, compile the appropriate …
- Distinctive: line 55
Run behavioral and economic lanes before increasing autonomy or fan-out.
- Distinctive: line 63
Every role owns a distinct output or policy boundary.
≈ “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
System Forge
Outcome
Design a bounded system whose topology, authority, shared state, and proof obligations are visible before execution, then compile it when a Foundry runtime is present.
Choose the smallest system shape
- Skill composition: one actor can perform the work sequentially.
- Swarm: independent or specialized workers need explicit coordination.
- Vertical: domain rules, audiences, taste, and evaluation recur across several capabilities.
- Plugin: a set of validated skills and optional MCP connectivity must be distributed to ChatGPT and Codex.
Do not create a swarm merely to make a workflow sound advanced. Parallelism is justified only for genuinely separable work, independent evaluation, latency reduction, or permission isolation.
Procedure
- Start with a Task Envelope.
- Include deliverables, stakes, autonomy, capability selection, permissions, required evidence lanes, and deployment targets.
- Resolve capabilities.
- Prefer registered skills and agents.
- Treat lexical graph suggestions as discovery hints, never silent routing decisions.
- Choose topology.
manager-workers: one owner decomposes and synthesizes.handoff: ownership moves between specialists.agents-as-tools: one agent retains control and invokes bounded specialists.parallel-merge: separable outputs join under a deterministic merge.debate-synthesize: independent hypotheses are judged and synthesized.
- Define authority and state.
- Give every role non-overlapping decision rights.
- Choose manager-only, append-only, or partitioned shared-state writes.
- Name the conflict owner and evidence-based resolution method.
- Define termination before dispatch.
- Write success conditions, stop conditions, maximum rounds, cost or latency ceilings, and escalation points.
- Keep builder, critic, and verifier separate when stakes or ambiguity justify it.
Select the execution mode.
- Foundry runtime available: only when
tools/foundry/cli.mjsandfoundry/contracts/exist in the current workspace, compile the appropriate pack:
node tools/foundry/cli.mjs forge \ --envelope <task-envelope.json> \ --pack <swarm-pack.json|vertical-pack.json|plugin-pack.json> \ --out <package-directory>- Portable skills-only install: read
references/portable-contracts.md, then author the Task Envelope and chosen pack with those field names. Mark compilation and proofpending-runtime; do not invent emitted host artifacts or an Evidence Receipt. Provide the command above as an optional handoff for a later SIS workspace run.
- Foundry runtime available: only when
- Prove static structure before live orchestration.
- Validate topology, capabilities, permissions, state policy, termination, and plugin shape.
- Run behavioral and economic lanes before increasing autonomy or fan-out.
Plugin rule
A skills-only plugin is a valid first release. Add an MCP server only when a real deployed endpoint, authentication model, privacy contract, and tool-level tests exist. Never invent a future endpoint in the manifest.
Quality gates
- Every role owns a distinct output or policy boundary.
- The synthesizer is named.
- Conflict resolution is explicit.
- Shared-state writes cannot silently overwrite another role.
- Termination and budget ceilings are machine-readable.
- Deployment claims match surfaces that were actually packaged and tested.
Return
Return the system-shape decision, topology, capability resolution, pack, execution mode, and the smallest blocked next step. Include compiled artifacts and an Evidence Receipt only when they were actually produced.
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.

