Route broad or ambiguous work across Simplicio Mapper, Fast, Dev CLI, Loop, and Runtime. Use when a request spans components, requires choosing the correct capability, needs an end-to-end workflow, or the agent is unsure which Simplicio skill to invoke. Prism classifies and composes; it does not execute mutations itself.
Scanned 9/5/2026
Install to Claude Code
npx -y skills add wesleysimplicio/simplicio-loop --skill simplicio-prism --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Simplicio Prism?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/wesleysimplicio-simplicio-prism)More formats (shields.io, HTML) on the badges page.
---
name: simplicio-prism
description: Route broad or ambiguous work across Simplicio Mapper, Fast, Dev CLI, Loop, and Runtime. Use when a request spans components, requires choosing the correct capability, needs an end-to-end workflow, or the agent is unsure which Simplicio skill to invoke. Prism classifies and composes; it does not execute mutations itself.
---
# Simplicio Prism
## Worker preflight and centralized artifact policy
Before routing or operating, each worker reads repository `AGENTS.md`, `CLAUDE.md`, and all relevant local skills. Prism only composes a route after that preflight. One binary/artifact set is built centrally from the canonical default branch and shared read-only; workers never rebuild binaries or regenerate canonical Mapper/Fast artifacts. Worktrees isolate source edits and receipts only. Route evidence and receipts carry repository/revision, binary digest/version, Mapper generation, and artifact digest. Missing, stale, incompatible, or mismatched central artifacts fail closed and select the central rebuild path only.
The async boundary is explicit: Loop/Prism uses Python `asyncio` for scheduling, leases, and I/O; Runtime/Tokio owns gates, effects, receipts, and reconciliation. Neither async layer authorizes a worker-local rebuild or canonical artifact regeneration.
Use Prism as the top-level capability router. Read `references/capabilities.yaml` for routing rules and `references/recipes.md` for end-to-end compositions. Load a component skill only after Prism has selected it; do not duplicate component documentation here.
## Generate the complete interface map
Run the inventory generator against the actual checkout before relying on an interface:
```bash
python3 scripts/generate_capability_inventory.py /path/to/repository \
--output /path/to/repository/capability-inventory.json
python3 scripts/generate_capability_inventory.py \
--validate /path/to/repository/capability-inventory.json
```
The inventory covers CLI entry points and subcommands, MCP registrations, public Python and Rust APIs, configuration files, inferred inputs/outputs/effects, observed errors and fallbacks, dependency/version data, compatibility, and cost estimates. Static cost and semantic contracts remain `requires_review` until measured or supplied in `capability-overrides.json`. Use `references/capability-record.schema.json` and `references/discovery.md` to interpret confidence.
## Routing algorithm
1. Classify intent as survey, retrieve, mutate, validate, orchestrate, govern, or mixed.
2. Check repository, revision, scope, availability, preconditions, and side-effect policy.
3. Select the smallest capability set and order dependencies before dependents.
4. Require Mapper before non-trivial mutation; require Dev CLI for mutation.
5. Add Fast for compatible repeated/large retrieval.
6. Add Loop for multi-step, parallel, retryable, or convergent work.
7. Add Runtime only for MCP, native execution, gates, receipts, checkpoints, backpressure, or governed subagents.
8. Emit a routing decision with reasons, fallbacks, and expected evidence.
For task-count routing, one to three tasks use direct parallelism; more than three tasks activate
Prism. An omitted quantity defaults to a minimum logical batch of ten tasks per slot. Slot count,
slot capacity, and Prism wave width have no logical upper bound; physical resource/lease governors
may still defer execution. Always query Fast from the fresh Mapper snapshot before selecting the
route when the repository already provides that integration.
## Non-negotiable boundaries
- Prism never edits files and never fabricates a capability.
- Unknown or unavailable capabilities become explicit `unresolved` items.
- Loop remains usable without Runtime.
- A reported success is not completion until the selected verifier proves it.
- Prefer the cheapest capability that satisfies the contract.
## Contract
Return `route_id`, `intent`, `selected_capabilities`, `order`, `preconditions`, `fallbacks`, `cost_estimate`, `evidence_requirements`, and `unresolved`.
## Resources
- `references/capabilities.yaml`: cross-component registry and decision matrix.
- `references/recipes.md`: tested compositions for common workflows.
- `scripts/probe-capabilities.py`: validate the registry and detect duplicate IDs.
## Runtime precedence (normative, 2026-08-05)
Consult Runtime before composing a cross-component route. Runtime has peer routes
for Mapper, Fast, and Loop only; Dev CLI is selected only as `loop.dev_cli` after
Loop activation. Direct file edits default to `simplicio edit`. Standalone
component packages remain callable when Runtime or a flow is absent.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!