Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/get-tmonier/argot/semantic-contractgit clone --depth 1 https://github.com/get-tmonier/argotWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/get-tmonier/argot/semantic-contract)<a href="https://agentmods.dev/agents/get-tmonier/argot/semantic-contract"><img src="https://agentmods.dev/badge/agents/get-tmonier/argot/semantic-contract.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01545 |
| Opus 5 | $0.00000 | $0.00772 |
| Sonnet 5 | $0.00000 | $0.00309 |
| Haiku 4.5 | $0.00000 | $0.00154 |
Grade A, and why
semantic-contract scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.
The semantic layer's self-calibration contract
Companion to calibration-contract.md, which governs
the base statistical scorer. The semantic layer (feature semantic, crate
crates/argot-rules-semantic/) calibrates itself separately at fit time; this
file records the invariants a contributor must not break.
The embedder
A static token-embedding table compiled into the binary — 15.6 MB of int8
weights under crates/argot-rules-semantic/model/, distilled from
jina-embeddings-v2-base-code with the model2vec technique, read by argot's own
forty lines of inference in static_embedder.rs. Embedding is a table lookup,
a sum in token order, an L2 normalise, and an f16 canonicalisation.
Three consequences a contributor should hold onto:
- There is no download, no cache to warm, no accelerator. A fit works air-gapped on the first run. Any code path that reintroduces a network dependency on the analysis side breaks this.
- Pooling order is part of the determinism contract. The sum runs in token
order over a fixed row layout and the result is canonicalised through f16, so
a cached vector and a freshly computed one are bit-identical. That identity is
what lets
~/.cache/argot/embeddings/be a pure accelerator. - It cannot represent order or structure. A bag of token vectors is a weaker sense than the transformer it replaced. The layer is viable because the structural confirmation below carries the precision — not because the embedding is strong. Weakening that confirmation to "let the embedder decide" is the change most likely to quietly wreck this layer.
ARGOT_STATIC_MODEL=<dir> points the embedder at another model directory
(model.safetensors + tokenizer.json). It exists so a candidate model can be
swept before it is shipped; it is not a user knob and is not documented as one.
The shape
One artifact, .argot/semantic-index.json (SemanticArtifact), built at fit:
per language, every corpus function's embedding (int8 codes with one shared
dequantisation scale, base64) plus two structural fingerprints per function —
its callees and its IDF-weighted identifier subtokens — and two
self-calibrated config blocks:
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 119 lines · 0 tokens per session scan A ed80f990f9ab
semantic-contract is an agent published in the GitHub repository get-tmonier/argot (48 stars, last pushed 21d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,545 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
harness-doctor
Check a claudeconf-generated harness against its RECORDED constitution profile — advisory, evidence-driven, read-only. Use when the user asks to check, diagnose, or "doctor" their harness, verify pins/wiring/drift, see whether branch protection matches the enforcement checklist, or whether a profile upgrade is…
github-action-reviewer
Reviews GitHub Action composite action, shell scripts, jq filters, PR annotations, comments, and review integration.
ring:qa
Senior QA Analyst for financial systems. Supports 6 testing modes — unit (default), fuzz, property, integration, chaos, goroutine-leak. Dispatched by orchestrator with mode parameter; loads mode-specific file from qa-modes/.
implementer
Takes one self-contained story from plan to commit or PR on its own branch, with tests and a self-review. Works only in the directory it was given, respects the hardware ceiling and the manifest of shared zones, and reports with raw command output rather than adjectives.
lead
Workflow orchestrator. Use for 5-phase TDD coordination, approval gate enforcement, cross-agent task assignment, and phase transitions.
replanner
Triggered by failure-classifier on F2-F4 escalations. Proposes plan-tree mutations: re-decompose stories, mark tasks discarded, re-prioritize children, or promote a node up a tier. Read-only on code; mutations applied via master-planner.