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 skills add autonomous-ai/autonomous-workshop --skill cadgit clone --depth 1 https://github.com/autonomous-ai/autonomous-workshopWrote 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/skills/autonomous-ai/autonomous-workshop/cad)<a href="https://agentmods.dev/skills/autonomous-ai/autonomous-workshop/cad"><img src="https://agentmods.dev/badge/skills/autonomous-ai/autonomous-workshop/cad/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/autonomous-ai/autonomous-workshop/cad"><img src="https://agentmods.dev/badge/skills/autonomous-ai/autonomous-workshop/cad.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Memory Poisoning · line 183 Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.Fix: Protect agent memory and state from modification by untrusted content. Use read-only memory for critical instructions and validate all state changes.
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 5.1 | $0.00099 | $0.10692 |
| Opus 5.5 | $0.00040 | $0.04277 |
| Sonnet 5.5 | $0.00020 | $0.02138 |
| Haiku 4.5 | $0.00010 | $0.01069 |
Grade A, and why
cad 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 2d 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 — 386 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Motion verification is opt-in. Standalone make_round and verify_project
default to false; pass --check-motion true to enable it. Inside Workshop,
read the immutable run-root MAKE-OPTIONS.json: the tools inherit its
check_motion value and reject contradictory flags. A missing options file
in an older materialized run retains its mandatory motion policy. When false,
skip motion sweeps and required animation/reconstruction/review, even if a
manifest or assembly claims exist. Motion is unverified, never passed; do not
claim assemblability or working motion from skipped evidence. Build, fit,
print gates and still-image review remain required. These rules take
precedence over motion-specific requirements in references and templates.
A correction carries unchanged parts forward. This is the default for
workshop fix, recorded as carry_unchanged in the run-root
MAKE-OPTIONS.json under schema_version: 2;
motion_policy.carry_unchanged() reads it. It is false for every run with no
source to carry from, and for every run created before the policy existed --
those have no schema-2 document, so a refreshed tool still reads them the way
they were created. workshop fix --full declines it. The policy changes only
what a round may CARRY FORWARD. It lowers no threshold, relaxes no gate and
skips nothing about the assembly.
A seeded run records its Approved Proof. workshop wish|fix --seed-source
(Workshop ADR 0096) freezes seed: {approval_sha256, manifest_sha256} under
schema_version: 3, beside check_motion and, for a correction,
carry_unchanged: true; motion_policy.seed() reads it and is None for every
other run. It changes no gate.
What it permits, and only after the changed-part set is established by
building and exporting EVERY part and hashing each against the source
archive's make/made.json product_manifest:
- A part whose STEP is byte-identical to the source's needs no new isolated component round. The source archive's round history for that part is still exactly true of it, because a gate is a pure function of the STEP it reads. Carry that history forward and record the hash it is carried on.
- Its per-part measure reports are carried forward the same way, unchanged.
You are not trusted on this, and do not need to be: make_round --require-component-passes rebuilds every part and refuses assembly review
for any whose digest no longer matches its recorded pass. A part that moved
cannot be carried even if you try.
What the policy never touches:
- The hash proof itself. It is the whole warrant for everything above, so it is computed fresh, over every part, every time.
refs:assembly,validate:assemblyandinterfere:assembly. The assembly changes whenever any part does, and interference is a property of the whole set, never of a part.- Every gate on every CHANGED part, and the assembled-object rounds.
verify_project, the whole-set renders and the independent blind review. The final sweep is what would catch a mistake in the carry-forward reasoning, so scoping it would remove the one check that makes the rest safe.
Write a measure/quick-fix-carry.md naming every part carried forward, the
sha256 it was carried on, and what was regenerated instead. A reader must be
able to tell a carried report from a fresh one without diffing, because the
round record itself cannot say which run produced it. If the changed-part set
turns out to be most of the project, say so and regenerate everything: the
policy saves nothing there and the claim is weaker.
CAD generation, inspection, and validation
Provenance: maintained in earthtojake/text-to-cad. Use the installed local skill files as the runtime source of truth; the repository link is only for provenance and release review.
What ships with it
60 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 205 B
- LICENSE 1.0 KB
- references/bought-parts.md 13 KB
- references/build123d-modeling.md 14 KB
- references/cad-brief.md 6.4 KB
- references/image-derived-verification.md 5.4 KB
- references/inspection-and-validation.md 14 KB
- references/motion-manifests.md 21 KB
- references/motion-presentation.md 6.5 KB
- references/organic-lofts.md 1.9 KB
- references/parameters.md 7.9 KB
- references/positioning.md 13 KB
- references/print-optimisation.md 10 KB
- references/project-structure.md 15 KB
- references/repair-loop.md 5.6 KB
- references/run-cost.md 13 KB
- references/standard-elements.md 11 KB
- references/step-generation.md 19 KB
- requirements.txt 170 B
- scripts/artifact/__init__.py 81 B runs code
- scripts/artifact/__main__.py 1.7 KB runs code
- scripts/artifact/cli.py 4.1 KB runs code
- scripts/brepbundle.py 25 KB runs code
- scripts/cadcache.py 11 KB runs code
- scripts/cadfilament.py 16 KB runs code
- scripts/cadfits.py 11 KB runs code
- scripts/cadgen_daemon/__init__.py 57 B runs code
- scripts/cadgen_daemon/__main__.py 351 B runs code
- scripts/cadgen_daemon/client.py 9.6 KB runs code
- scripts/cadgen_daemon/server.py 15 KB runs code
- scripts/cadgen_daemon/watchdog.py 703 B runs code
- scripts/cadmount.py 31 KB runs code
- scripts/cadprint.py 7.5 KB runs code
- scripts/check_fit 13 KB
- scripts/check_layout 15 KB
- scripts/check_mesh 8.7 KB
- scripts/check_motion 89 KB
- scripts/check_mount 27 KB
- scripts/check_overhang 33 KB
- scripts/check_spec_format 19 KB
- scripts/check_spec_numbers 13 KB
- scripts/check_thickness 36 KB
- scripts/gen/__init__.py 67 B runs code
- scripts/gen/__main__.py 1.7 KB runs code
- scripts/gen/cli.py 6.4 KB runs code
- scripts/geometry_disclosure.py 6.1 KB runs code
- scripts/geometry_sources.py 9.0 KB runs code
- scripts/inspect/__main__.py 1.6 KB runs code
- scripts/inspect/inspect_refs/__init__.py 45 B runs code
- scripts/inspect/inspect_refs/cli.py 35 KB runs code
- scripts/inspect/inspect_refs/inspect.py 39 KB runs code
- scripts/inspect/inspect_refs/supervisor.py 8.9 KB runs code
- scripts/inspect/inspect_refs/worker_guard.py 1.3 KB runs code
- scripts/meshlib.py 20 KB runs code
- scripts/motion_parallel.py 7.6 KB runs code
- scripts/motion_policy.py 5.2 KB runs code
- scripts/motion_presentation.py 18 KB runs code
- scripts/motion_states.py 18 KB runs code
- scripts/packages/cadgen/LICENSE 1.0 KB
- scripts/packages/cadgen/pyproject.toml 1.4 KB
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.
- 2d ago Changed · +1 lines 7fc874327a94
- 3d ago Changed · +6 lines 14881e796d06
- 7d ago Changed · +1 lines · -11 tokens per session 2778c5adacc1
- 1mo ago Changed · +1 lines b0a9ad9a29d8
- 1mo ago Changed · +24 lines ab55a98a3c61
- 1mo ago Changed · +46 lines fac23ecc5e1d
- 1mo ago Changed · +11 lines 678dfd6f50fe
- 1mo ago Changed · +12 lines 7c9f76764eed
- 1mo ago Changed · +20 lines b88b89c0b70f
- 1mo ago First seen · 264 lines · 110 tokens per session scan A d05e3473f862
cad is a skill published in the GitHub repository autonomous-ai/autonomous-workshop (24 stars, last pushed yesterday), licensed Apache-2.0. It adds 99 tokens to every session and 10,692 once invoked, about $0.0004 per session on Opus 5.5. 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.
cad compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| cadautonomous-ai/autonomous-workshop | Skill | 24 | yesterday | Apache-2.0 | A | 99 | Codex |
| pennylaneK-Dense-AI/scientific-agent-skills | Skill | 56k | 4d ago | MIT | A | 74 | Claude Code |
| holoscan-install-wheelNVIDIA/skills | Skill | 3.9k | yesterday | Apache-2.0 | A | 34 | Claude Code, Codex |
| future-self-liteTHU-MAIC/OpenMAIC | Skill | 43k | yesterday | MIT | A | 20 | Claude Code, Codex |
| rocm-kernelshuggingface/kernels | Skill | 783 | yesterday | Apache-2.0 | A | 93 | Claude Code |
| typing-exclusion-workergetsentry/skills | Skill | 1.1k | yesterday | Apache-2.0 | A | 57 | Claude Code |
| precisemicroprediction/precise | Skill | 338 | 5d ago | MIT | A | 76 | Claude Code, Codex |