Menü
Authority declaration for framework self-build (target=Framework): owns the framework surface so guard passes framework writes, never product (I11). Use on a target=Framework wave.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
| name | framework-maintenance |
| description | Authority declaration for framework self-build (target=Framework): owns the framework surface so guard passes framework writes, never product (I11). Use on a target=Framework wave. |
The authority that lets the agent build the framework itself. When a task targets the Framework
(this rebuild, a ref-adopt, foundation hardening), its waves write to engine/ and library/ — paths
no kernel/support/system module owns. Without a declared authority, guard Layer 4 (I5) blocks those
writes. This module is that declaration: its authority.owns covers the framework surface, so guard
passes framework writes — and only framework writes (I11: never product space).
It is a declaration, not an orchestrator. It does no work and produces no receipts of its own — the building agent does the work and writes the milestone receipts (I10) under this authority.
Does: declare authority over the framework surface (engine/*, library/*, the registry/config/version
files) in skill-rules.json so guard-check.py authority --module framework-maintenance authorizes a
framework wave's write targets; carry the one-time bootstrap-attestation exception (Wave 1). Does not:
execute or transform anything (it is an authority grant), write to product space (I11 — that is a
hard boundary, not a default), or write its own receipts (the building agent's milestone receipts cover
the work).
scope.md.target: Framework — there is no framework authority to grant.target: Framework).registry.yaml (guard resolves this module's path → skill-rules.json).engine/shared/references/invariants.md (I5 authority, I11 space boundary).target: Framework task routes its waves through guard with --module framework-maintenance.skill-rules.json authority.owns and fnmatches every declared
write target against it (guard-check.py authority). All within owns → AUTHORIZED; any target
outside → UNAUTHORIZED (HARD error) — that target is either product space (forbidden, I11) or a new
framework path that must be added to owns deliberately, never silently passed.Produces no receipts itself — it is an authority declaration. It enables guard to authorize the
building agent's framework writes; the agent writes the milestone receipts (module: framework-maintenance,
e.g. these rebuild receipts) under this authority (I10). The contract is skill-rules.json authority.owns
schemas/framework-authority.schema.json (authority-scope + bootstrap-attestation record shape).authority.owns: do not widen owns reflexively —
first decide whether the target is product space (then it is an I11 violation, not an authority gap) or
a genuinely new framework path (then add it to owns as a deliberate, reviewed change).target: Framework: HARD stop (I11). Framework and product are
separate spaces; a task that needs both is two tasks.Done when: a target: Framework wave's write targets all fall within authority.owns and guard
Layer 4 returns AUTHORIZED (or Wave-1 bootstrap attestation is recorded). The building agent then
proceeds and writes its own receipt.
Not done if: a product-space write was authorized under this module (I11 breach); owns was widened
to pass a wave without deciding framework-vs-product; the module tried to do work or write a receipt
itself.
Next step: executor runs the authorized framework wave; the building agent writes the milestone receipt.
authority.owns at Layer 4. framework-maintenance declares; guard enforces.Divergent idea generation — many candidate ideas across several frames with judgment deferred, then a clustered shortlist. Use to open a solution space wide, or /brainstorm. Not for developing one idea (ideate) or picking one (propose).
Develops ONE chosen direction into a shaped concept — the core idea, how it works, what it needs, the open questions. Use to deepen a single idea, or /ideate. Not for generating many (brainstorm) or deciding among them (propose).
Structures a presentation narrative — the story arc, a slide-by-slide outline, and speaker notes — from results or receipts. Use to shape a talk or deck, or /present. Not for prose reports (report) or rendering visuals (asset).
Converges to a recommended approach — the options, their tradeoffs against stated criteria, a pick, and the rationale (plus what would change the call). Use to decide among approaches, or /propose. Not for generating options (brainstorm) or deep-diving one (ideate).
Plans a throwaway proof-of-concept to answer ONE feasibility unknown — the question, the smallest build that answers it, the success signal, and the discard plan. Use to de-risk before committing, or /prototype. Not for the real build (executor) or deciding (propose).
Produces a REDTEAM-FINDINGS.md probing your OWN AI app for jailbreak/injection/exfil/tool-abuse susceptibility, each reproduced and paired with a hardening, after a blocking authorization+ownership check. Use to red-team an owned AI system defensively.