hex-core
Skill15 starsChanged 18 days agoApache-2.0View on GitHub
Shared reference library for the hex swarm-orchestration bundle - worker role registry, model capability matrix, swarm protocol vocabulary, and the .agents/memory/hex.md memory spec. Not invoked directly; the hex-plan, hex-execute, hex-review, hex-architect, and hex-init skills link into it.
In this repository45 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .agents/research/hooks_vendor_reports/claude.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- code-docs-cleanupSkill
- docs-instrumentSkill
- docs-planSkill
- docs-reviewSkill
- hex-architectSkill
- hex-discussSkill
- hex-executeSkill
- hex-finalizeSkill
- hex-initSkill
- hex-loopSkill
- hex-planSkill
- hex-retroSkill
- hex-reviewSkill
- nox-reviewSkill
- bugfixSkill
- builderSkill
- changelogSkill
- code-checkSkill
- code-docs-cleanupSkill
- commitSkill
- docs-instrumentSkill
- docs-planSkill
- docs-reviewSkill
- docsSkill
- finalizeSkill
- hex-architectSkill
- hex-coreSkill
- hex-discussSkill
- hex-executeSkill
- hex-finalizeSkill
- hex-initSkill
- hex-loopSkill
- hex-planSkill
- hex-retroSkill
- hex-reviewSkill
- meta-maintain-configSkill
- meta-validate-contextSkill
- nextSkill
- nox-reviewSkill
- qa-engineerSkill
- security-auditorSkill
- upstream-refreshSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
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 41
Every hex skill installs as a sibling under the client's skills directory, so references resolve by relative path:
- Very specific: line 48
If hex-core is not installed, add it:
≈ “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
hex-core — Shared Reference Library
hex-core is the library skill of the hex swarm bundle. It holds the
vocabulary and contracts every other hex skill builds on, kept in one
place so they never drift apart. It is never invoked directly — there
is no orchestration flow here and no $ARGUMENTS. The orchestrator skills
(/hex-plan, /hex-execute, /hex-review, /hex-architect) and
/hex-init link into these references.
References
| Reference | Holds |
references/protocol.md | The bundle's spine — shared shape, tier and overlay grammar, the meta-plan approval gate, spawn-selection precedence, worker coordination (workers never wait, the build-job budget, event-driven completion, degraded modes), traceability IDs, untrusted-text echoes, the constitution gate, the handoff contract, and the per-run upkeep step: what every mode needs. Also carries the load map (which mode opens which sibling beyond the spine) and the what-moved-where pointer table into the six sibling files below, which now hold the per-topic contracts protocol.md used to carry directly. |
references/loop.md | The canonical Review-Fix Loop (### The last-reviewed anchor, ### Anchor validation, ### Delta round scope, ### The diminishing-returns stop) and the Convergence contract. Conditional-load — see protocol.md's load map for which modes open it. |
references/decompose.md | Parallel-by-default decomposition — pipelines of steps cut along contracts, the contract wave, the Parallelization table, the schedule log. Conditional-load — see protocol.md's load map for which modes open it. |
references/worktree.md | Pipeline worktree mechanics. Conditional-load — see protocol.md's load map for which modes open it. |
references/verify.md | Verification. Conditional-load — see protocol.md's load map for which modes open it. |
references/adversary.md | The adversary contract. Conditional-load — opened only when adversary=on; see protocol.md's load map. |
references/severity.md | Finding severity. Conditional-load — see protocol.md's load map for which modes open it. |
references/workers.md | The worker role index plus the universal worker protocol; full personas (mission, spawn-prompt template, focus modes, output contract) live one file per role under references/workers/ — explorer, architecture-explorer, researcher, builder, tester, reviewer, doc-reviewer, architect, simulator. Orchestrators load only the personas in the resolved spawn set. Workers are prompt blocks the orchestrator copies into a subagent, not shipped agent files. |
references/models.md | The one model matrix for the whole bundle: the four capability classes (light, standard, standard-high, deep) and the recommended class per role, plus the escalation and instantiation rules. No model guidance lives anywhere else in hex. |
references/memory.md | The .agents/memory/hex.md memory spec: the three sections and their per-section ownership, resolution order, a complete example file, plus the editing, destination-choice, and staleness rules. |
references/config.md | The hex.md › Preferences config vocabulary — keys, defaults, merge rules. Conditional-load — read only when hex.md › Preferences contains a fenced yaml block. |
references/archive.md | The spec fold-back mechanics — delta grammar, destination resolution, the safety envelope and its four commands, halt semantics, idempotence and the fold receipt, revert, and the plan archive. Sole definition site; hex-execute, hex-review, and protocol.md link here. Conditional-load — read only when the plan under review carries a ## Spec Deltas block. |
references/finalize.md | The remote-rights boundary — the act set and its branch scoping, the consent model, the literal force-push form, the backup-ref armed/inert lifecycle, remote verification and its ceilings, re-entry, the degrade ladder, the trust classes, the pre-flight halt texts, and the placeholder-substitution rule. Sole definition site; every bundle-wide remote-rights qualifier links here. Conditional-load — read only when finalizing a branch or resolving a remote-rights qualifier. |
references/resources.md | The resource knob sheet — the job budget, gate-only tools, never-wait, worktree cleanup, and the per-ecosystem knob sheet. Sole definition site; hex never defines how to verify a project. Conditional-load — read only when a run will build or test. |
How sibling skills reference this
Every hex skill installs as a sibling under the client's skills directory, so references resolve by relative path:
See [`../hex-core/references/protocol.md`](../hex-core/references/protocol.md).If hex-core is not installed, add it:
grim add ghcr.io/michael-herwig/arcana/hex-core:latestCross-skill mentions elsewhere use the command form: /hex-init,
/hex-discuss, /hex-loop, /hex-plan, /hex-execute, /hex-review,
/hex-architect, /hex-finalize.
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.