Semantic ISA
Separate model proposals from executable authority. Translate a named agent's
capabilities into a closed instruction set, then specify what a deterministic
admission function permits from a captured policy and evidence snapshot.
This skill designs that contract; loading it installs no runtime enforcement,
grants no tools and authorises no external action.
Choose DESIGN for a system contract or WRITE for an explanation. Read
research context before attributing claims or numbers
to the paper. Use the specification template for
DESIGN. Read Codex tool integration when the system
uses this Mac's connectors, memory, browser, MCP or recurring work.
DESIGN
Read the repository's instructions, architecture and relevant runtime
contracts. Inventory capabilities from actual tools, source and authorised
configuration. Distinguish available, configured, granted, proposed and
demonstrated. Ask only for consequential missing decisions; do not ask the
user to recite tools that can be inspected.
List side effects and data disclosure: reads can expose private content;
simulations can incur model cost; browser navigation can send data. Include
storage, network, messages, paid queries, delegation and scheduled work.
Name what is outside the requested agent and cannot be reached by its kernel.
Map capabilities to domain-specific instructions. Avoid an unrestricted
tool_call(name, args) escape hatch. Classify instructions using the five
cores below, or document an explicit domain adaptation. Every operation needs
a governance property, binding, authority requirement and error contract.
| Core |
Role |
Boundary |
| Cognitive |
Generate, decompose, reflect |
Untrusted proposals, never execution authority |
| Memory |
Load, store, transform, render |
Preserve lineage; lossy/model transforms are not deterministic facts |
| Execution |
Domain writes, external tools, delegation, responses |
Check each effect and disclosure destination |
| Normative |
Verify, constrain, refuse, interrupt |
Trusted checks; models may request but cannot issue their own proof |
| Meta-cognitive |
Progress estimates, routing, resource observation |
Advice cannot increase grants or budgets |
Define closed input/output schemas: units, enums, bounds, unknowns, resource
identities, revisions and version negotiation. Reject malformed values;
any legitimate normalization must be explicit, deterministic and auditable.
Keep retrieved instructions as data. Typed syntax alone proves no factual
truth, user intent or right to act.
Build the Security Context Registry and Instruction Dependency Graph.
Track provenance, confidentiality, integrity/review state and authority as
separate dimensions. Trace tags through storage, caching, rendering and
delegation. For each sink, name its required checks and accepted evidence.
Specify VERIFY as a named deterministic predicate over captured inputs,
policy, principal, resource, destination, revision and expiry. Its receipt
discharges only that predicate for those exact inputs. Preserve original
provenance and privacy labels. Never implement a universal trusted = true
after a schema check or model review. External facts may remain unknown.
Define structured refusals and recovery. Include stale revisions, revoked
authority, conflicting idempotency keys, uncertain remote writes and resource
limits. Reconciliation precedes retry when a side effect may already exist.
No silent coercion, invented success or automatic extra spend.
Draw container and sequence diagrams. Trace one refused sink, a bounded
verification, then the permitted or still-refused outcome. Distinguish a
reproducible admission decision from nondeterministic external execution.
Plan deterministic fixtures: one refusal/success pair per sink, failed
verification, provenance round trip, replay, concurrency and recovery.
Drive scripted proposals, not a live model. Test the real write boundary;
a diagram or a green schema validator is not enforcement proof.
Keep the pure admission predicate and domain types free of I/O, ambient clocks
and model calls. The application gathers current authenticated state, captures
an explicit as-of time, runs admission and rechecks volatile authority at commit.
Infrastructure executes effects and persists receipts. TOCTOU protection,
durable reservations and destination readback belong in that execution contract.
Compose other skills only when useful and installed: brainstorming for product
scope, arch for diagrams, rust-clean-architecture for placement,
security-review for sink analysis, and integration-test for executable proof.
If implementation planning is requested, use codex-prd and codex-beads with
the actual tracker (cn --toon on this Mac). Missing optional skills are not a
reason to invent tools or stop a specification that can be completed directly.
WRITE
Describe Arbiter-K as a 2026 research proposal/prototype. Its hardware analogy
specifies an execution boundary; it does not make an LLM or remote service
deterministic. Distinguish the paper's vocabulary from your own application ISA.
Read original sources before quoting figures; include benchmark conditions,
residual failures and limitations. If a source cannot be checked, omit its
numbers or mark them unverified. Do not turn a secondary critique into a fact
about the implementation. Follow the user's writing/voice guidance when present.
Completion and rollout
Every in-scope capability maps to an instruction and every sink has its schema,
policy, proof requirements, refusal and recovery tests. List uncovered surfaces
and unresolved authority rather than claiming the whole environment is governed.
Save the design in the target repository's indexed docs/ layout. Commit or
publish only with the user's existing authorisation and repository rules.
A proposed kernel can run shadow evaluation on captured fixtures or beside
existing controls. Shadow mode never bypasses those controls or enables effects
that current policy refuses. Enforce only after reviewing refusal traces and
proving the actual executor cannot route around admission. Documentation and
skill installation alone do not meet that bar.
1---2name: semantic-isa3description: Design a typed Semantic Instruction Set Architecture for an agent or MCP workflow, with deterministic admission, provenance, scoped verification and refusal tests; or explain Arbiter-K accurately. Use for semantic ISA, agent execution kernels, tool-call taint tracking and governance contracts, rather than ordinary API design or a general security audit.4metadata:5 source: https://skillmd.com/skills/decebal/semantic-isa6---78# Semantic ISA910Separate model proposals from executable authority. Translate a named agent's11capabilities into a closed instruction set, then specify what a deterministic12admission function permits from a captured policy and evidence snapshot.13This skill designs that contract; loading it installs no runtime enforcement,14grants no tools and authorises no external action.1516Choose **DESIGN** for a system contract or **WRITE** for an explanation. Read17[research context](references/arbiter-k.md) before attributing claims or numbers18to the paper. Use [the specification template](references/isa-template.md) for19DESIGN. Read [Codex tool integration](references/codex-tools.md) when the system20uses this Mac's connectors, memory, browser, MCP or recurring work.2122## DESIGN23241. Read the repository's instructions, architecture and relevant runtime25 contracts. Inventory capabilities from actual tools, source and authorised26 configuration. Distinguish available, configured, granted, proposed and27 demonstrated. Ask only for consequential missing decisions; do not ask the28 user to recite tools that can be inspected.292. List side effects and data disclosure: reads can expose private content;30 simulations can incur model cost; browser navigation can send data. Include31 storage, network, messages, paid queries, delegation and scheduled work.32 Name what is outside the requested agent and cannot be reached by its kernel.333. Map capabilities to domain-specific instructions. Avoid an unrestricted34 `tool_call(name, args)` escape hatch. Classify instructions using the five35 cores below, or document an explicit domain adaptation. Every operation needs36 a governance property, binding, authority requirement and error contract.3738 | Core | Role | Boundary |39 | --- | --- | --- |40 | Cognitive | Generate, decompose, reflect | Untrusted proposals, never execution authority |41 | Memory | Load, store, transform, render | Preserve lineage; lossy/model transforms are not deterministic facts |42 | Execution | Domain writes, external tools, delegation, responses | Check each effect and disclosure destination |43 | Normative | Verify, constrain, refuse, interrupt | Trusted checks; models may request but cannot issue their own proof |44 | Meta-cognitive | Progress estimates, routing, resource observation | Advice cannot increase grants or budgets |45464. Define closed input/output schemas: units, enums, bounds, unknowns, resource47 identities, revisions and version negotiation. Reject malformed values;48 any legitimate normalization must be explicit, deterministic and auditable.49 Keep retrieved instructions as data. Typed syntax alone proves no factual50 truth, user intent or right to act.515. Build the Security Context Registry and Instruction Dependency Graph.52 Track provenance, confidentiality, integrity/review state and authority as53 separate dimensions. Trace tags through storage, caching, rendering and54 delegation. For each sink, name its required checks and accepted evidence.556. Specify `VERIFY` as a named deterministic predicate over captured inputs,56 policy, principal, resource, destination, revision and expiry. Its receipt57 discharges only that predicate for those exact inputs. Preserve original58 provenance and privacy labels. Never implement a universal `trusted = true`59 after a schema check or model review. External facts may remain unknown.607. Define structured refusals and recovery. Include stale revisions, revoked61 authority, conflicting idempotency keys, uncertain remote writes and resource62 limits. Reconciliation precedes retry when a side effect may already exist.63 No silent coercion, invented success or automatic extra spend.648. Draw container and sequence diagrams. Trace one refused sink, a bounded65 verification, then the permitted or still-refused outcome. Distinguish a66 reproducible admission decision from nondeterministic external execution.679. Plan deterministic fixtures: one refusal/success pair per sink, failed68 verification, provenance round trip, replay, concurrency and recovery.69 Drive scripted proposals, not a live model. Test the real write boundary;70 a diagram or a green schema validator is not enforcement proof.7172Keep the pure admission predicate and domain types free of I/O, ambient clocks73and model calls. The application gathers current authenticated state, captures74an explicit as-of time, runs admission and rechecks volatile authority at commit.75Infrastructure executes effects and persists receipts. TOCTOU protection,76durable reservations and destination readback belong in that execution contract.7778Compose other skills only when useful and installed: `brainstorming` for product79scope, `arch` for diagrams, `rust-clean-architecture` for placement,80`security-review` for sink analysis, and `integration-test` for executable proof.81If implementation planning is requested, use `codex-prd` and `codex-beads` with82the actual tracker (`cn --toon` on this Mac). Missing optional skills are not a83reason to invent tools or stop a specification that can be completed directly.8485## WRITE8687Describe Arbiter-K as a 2026 research proposal/prototype. Its hardware analogy88specifies an execution boundary; it does not make an LLM or remote service89deterministic. Distinguish the paper's vocabulary from your own application ISA.90Read original sources before quoting figures; include benchmark conditions,91residual failures and limitations. If a source cannot be checked, omit its92numbers or mark them unverified. Do not turn a secondary critique into a fact93about the implementation. Follow the user's writing/voice guidance when present.9495## Completion and rollout9697Every in-scope capability maps to an instruction and every sink has its schema,98policy, proof requirements, refusal and recovery tests. List uncovered surfaces99and unresolved authority rather than claiming the whole environment is governed.100Save the design in the target repository's indexed `docs/` layout. Commit or101publish only with the user's existing authorisation and repository rules.102103A proposed kernel can run shadow evaluation on captured fixtures or beside104existing controls. Shadow mode never bypasses those controls or enables effects105that current policy refuses. Enforce only after reviewing refusal traces and106proving the actual executor cannot route around admission. Documentation and107skill installation alone do not meet that bar.