lihonggang-scnu/veritmm/veritmm-tmm

Claude Code and agent skill by lihonggang-scnu/veritmm

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Run verified multilayer-optics simulations through VeriTMM's verifier-first transfer-matrix protocol and receive a physics acceptance certificate. Use for thin films, DBRs, photonic crystals, defect cavities, chirped stacks, and absorbers — any passive, isotropic, planar 1-D multilayer task. Prefer the veritmm MCP tools; the veritmm CLI is the fallback. Every result is independently verifiable with verify_run.

Install

npx vskill@latest install lihonggang-scnu/veritmm/veritmm-tmm

SKILL.md

VeriTMM — verifier-first multilayer optics (TMM)

VeriTMM executes passive, isotropic, planar 1-D multilayer optics and issues a machine-readable physics acceptance certificate for every result. The division of labor is fixed: the agent proposes, VeriTMM computes, the deterministic verifier certifies. Neither you nor any optimizer can certify a result — verify_run re-derives the verdict from persisted evidence and policy.

Preferred surface: the veritmm MCP tools

Servers start in compact mode: only veritmm_discover and veritmm_catalog_get are bound; call veritmm_discover first and request the full grid for the complete namespace. Every tool description carries an effect tag — [READ_ONLY], [COMPUTATIONAL], or [EXPENSIVE_COMPUTE].

Namespaced grid (full mode; 15 tools, underscore style):

ToolTagPurpose
veritmm_discover[READ_ONLY]Compact discovery of the tool grid
veritmm_catalog_get[READ_ONLY]Runtime-generated capability catalog
veritmm_material_search / veritmm_material_explain[READ_ONLY]Bundled optical-constant catalog
veritmm_stack_validate[READ_ONLY]Validate a stack skeleton
veritmm_simulate_spectrum[COMPUTATIONAL]Managed simulate execution
veritmm_simulate_batch[EXPENSIVE_COMPUTE]Batch of isolated jobs + aggregate report
veritmm_gradient_compute[COMPUTATIONAL]Thickness gradients (proposal basis)
veritmm_sensitivity_analyze[COMPUTATIONAL]Normalized sensitivities (proposal basis)
veritmm_optimize_problem[EXPENSIVE_COMPUTE]Declarative problem -> optimize -> certify
veritmm_verify_run[COMPUTATIONAL]Independent four-state verification
veritmm_verify_explain[COMPUTATIONAL]Why is run A better than run B (provenance)
veritmm_job_status / veritmm_job_result[READ_ONLY]Job state and outputs
veritmm_intent_compile[COMPUTATIONAL]IntentSpec -> task + equivalence certificate

Legacy grid (still bound in full mode; run and verify_run are deprecated aliases of veritmm_simulate_spectrum / veritmm_verify_run):

ToolPurpose
describeCapability manifest + protocol identity scheme
schemaExport the exact task JSON Schema (kind: simulation)
examplesList bundled example tasks
preflightGate one task without running a spectrum
runManaged execution of one simulate task (default acceptance settings)
history / inspect / lineage / comparePersisted experiment state
verify_runIndependent four-state verification of a run

CLI fallback (same capabilities, same semantics): veritmm <command> --json. Large artifacts are MCP resources (veritmm://run/{run_id}/{path}); with the CLI they are files inside the run directory.

Call order

  1. describe — capability manifest and identity scheme.
  2. schema simulation — read the contract before writing task JSON.
  3. Compose ONE complete task: wavelengths and thicknesses in nanometres,

angles in degrees; select material datasets explicitly when identity matters.

  1. preflight — if rejected, follow the typed failure codes and safe actions;

never guess a repair and never work around a rejection.

  1. run — returns the compact envelope: status, certificate_id, artifact

references. Fetch details via resources only when the next decision needs them.

  1. verify_run — read the four states: integrity_status,

certification_status, replay_status, authenticity_status, and overall_status. A cache replay is certified with replay_status: unavailable — that is normal and not a failure.

Boundaries and red lines

  • Supported physics: passive, isotropic, planar 1-D multilayers (thin films,

DBRs, photonic crystals, defect cavities, chirped stacks, absorbers, finite substrates). Gratings, metasurfaces, anisotropy, nonlinear optics, finite beams, dipoles, mode sources, and time-domain requests fail closed with typed failures — do not attempt workarounds.

  • A = 1 - R - T is absorption in finite layers; E_system = 1 - R is a

different quantity with a different validity condition.

  • A successful run or a high score is not physics validity: the verdict

lives in the certificate and in verify_run.

  • Robustness and yield are separate artifacts; a physics certificate does not

imply robust optimality.

  • Never modify or re-hash artifacts inside a run directory.
  • run on this surface is simulate-mode only; optimization, sweeps,

sensitivity, and tolerance studies are operator-side CLI work outside this skill.

Details: references/protocol.md (tool surface and task format), references/artifacts.md (artifact sets and reading order), references/certificate.md (certificate fields and the four verify-run states), references/boundaries.md (capability boundary and material governance).

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
77
Repository forks
15
7-day trend score
22

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/lihonggang-scnu/veritmm/veritmm-tmm/badge)](https://verified-skill.com/skills/lihonggang-scnu/veritmm/veritmm-tmm)

Something wrong with this skill or its page? Tell us.