lihonggang-scnu/veritmm/veritmm-tmm
Claude Code and agent skill by lihonggang-scnu/veritmm
- community
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-tmmSKILL.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):
| Tool | Tag | Purpose | |
|---|---|---|---|
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):
| Tool | Purpose | |
|---|---|---|
describe | Capability manifest + protocol identity scheme | |
schema | Export the exact task JSON Schema (kind: simulation) | |
examples | List bundled example tasks | |
preflight | Gate one task without running a spectrum | |
run | Managed execution of one simulate task (default acceptance settings) | |
history / inspect / lineage / compare | Persisted experiment state | |
verify_run | Independent 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
describe— capability manifest and identity scheme.schema simulation— read the contract before writing task JSON.- Compose ONE complete task: wavelengths and thicknesses in nanometres,
angles in degrees; select material datasets explicitly when identity matters.
preflight— if rejected, follow the typed failure codes and safe actions;
never guess a repair and never work around a rejection.
run— returns the compact envelope:status,certificate_id, artifact
references. Fetch details via resources only when the next decision needs them.
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 - Tis absorption in finite layers;E_system = 1 - Ris 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.
runon 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
- SKILL.md in the repositoryCheckingtmm_engine/skills/veritmm-tmm/SKILL.md
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
[](https://verified-skill.com/skills/lihonggang-scnu/veritmm/veritmm-tmm)