Menú
Templates and risk-tier evolution policy for authoring new agents, skills, and hooks — the 3 artifact kinds this harness produces (no "mcp" kind; this harness has no MCP). Used primarily by meta-agent when extending the harness. Auto-invocable.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
Canonical model-tier reference and the Fable-5 manual-only enforcement protocol. Consult before writing or changing any model field (agent frontmatter, skill frontmatter, Task dispatch parameter, settings.json availableModels, env vars). Auto-invocable — any agent or skill authoring a new artifact should load this first.
Reference-only skill documenting where artifacts live, the 3-verdict vocabulary (pass/pass-with-notes/reject), and the Preserved-Invariants-vs-Changed-Behaviors contract for revising an existing artifact. Load before writing or revising any artifact under .fable/. Auto-invocable.
Show CONSTITUTION.md's content and invariant list on request, re-run the SHA-256 hash check against CONSTITUTION.sha256 on demand, and document the amendment procedure. Useful for Claude to consult proactively as well as for a user to invoke directly.
Merge-safe installer that makes FABLE-HARNESS's agents/skills/hooks available at Claude Code's user scope (~/.claude), so they work across every project, not just H:\FABLE-HARNESS. Manual-only — never auto-invoked, since this touches shared, hard-to-reverse state outside this project.
Highest-stakes planning skill using Fable-5. Manual-only — Claude can never auto-invoke this; only a user typing /plan-deep can. Use for the small number of asks big enough to warrant Fable-level structure and thinking budget (major-scope architecture, cross-cutting rewrites, irreversible decisions).
Everyday, auto-invocable planning skill for routine feature/bug/refactor work. Sonnet-tier — the non-Fable default. Use for most planning needs; escalate to /plan-deep (manual-only, Fable-5) only for the rare highest-stakes plan.
| name | agent-factory |
| description | Templates and risk-tier evolution policy for authoring new agents, skills, and hooks — the 3 artifact kinds this harness produces (no "mcp" kind; this harness has no MCP). Used primarily by meta-agent when extending the harness. Auto-invocable. |
| model | sonnet |
FABLE-HARNESS authors exactly three kinds of artifact: agent, skill, hook. (Sibling-repo research references a fourth "mcp" kind — not applicable here; CONSTITUTION.md N8 means this harness never registers an MCP server, so there is no mcp-config artifact kind to template.)
.claude/agents/<name>.md)---
name: <lowercase-hyphen-name>
description: <dense description including trigger conditions — this drives delegation>
tools: Read, Write, Edit, Glob, Grep, Bash
model: sonnet # or haiku / opus per model-routing skill's tier table — NEVER fable
---
Required: name, description. tools: should be an explicit allowlist built from Read/Write/Edit/Glob/Grep/Bash + Task dispatch only — no MCP tool names anywhere (N8). Consult the model-routing skill's tier table before picking model:.
.claude/skills/<name>/SKILL.md)---
name: <name>
description: <dense description — drives auto-invocation unless disable-model-invocation is set>
model: sonnet # optional; omit to inherit, or haiku/opus/sonnet per tier table
# disable-model-invocation: true # only for plan-deep and install-user-scope — manual-only skills
---
Required: name, description. Add disable-model-invocation: true only when the skill must be unreachable by Claude's own judgment (currently: plan-deep, install-user-scope). context: fork, allowed-tools, argument-hint are optional per the skills reference.
.claude/settings.json hooks.<EventName>[], plus the script it points at under .claude/hooks/){
"matcher": "Bash",
"hooks": [
{ "type": "command", "command": "pwsh -File .claude/hooks/pretooluse-bash-safety.ps1" }
]
}
Hook scripts must: fail open on internal error (exit 0 on unexpected parse/exception — N5), fail closed on a genuine policy violation (exit 2, never exit 1 — N5's explicit gotcha callout), and — if enforcing a whitelist (e.g. the Bash safety gate) — normalize input (strip quotes, collapse whitespace, split on &&/|/;) before matching each subcommand independently against an explicit allow-set (N6).
| risk_class | evolution_policy |
|---|---|
| low | Auto-commit the new artifact once it passes schema/invariant validation — no human gate required. |
| medium | Require a human review before the artifact is considered final; may be drafted and staged, but not promoted, without sign-off. |
| high | Require a human review before final; treat any ambiguity as a reason to escalate rather than assume. |
| critical | Require a human review before final, and prefer pairing it with an explicit decision record under memory/decisions/ regardless of outcome. |
meta-agent (or any authoring agent) must classify every new artifact's risk_class before deciding whether it can auto-commit. When in doubt, round up to the more conservative tier — this mirrors the evolution-handoff propose/evaluate/commit/HITL contract pattern from sibling-repo research, reimplemented here without any MCP/TheEights dependency (N8): "auto-commit" and "require human review" are just local git-commit-or-don't decisions, not calls to an external evolution service.
No agent file ever sets model: fable. This is non-negotiable and is CONSTITUTION.md N2. If meta-agent or any other authoring flow is about to write model: fable into anything other than .claude/skills/plan-deep/SKILL.md, it must stop and cite Refused per N2: no agent file may ever set model: fable. See the model-routing skill for the full six-surface reasoning behind why this rule exists.
N2 (Fable manual-only — the hard rule above), N7 (typed-agent provenance — never scaffold a "general-purpose" stand-in when a typed producer is what's needed), N8 (no MCP — hook/agent tool allowlists never include mcp__* names).