Stop "parts in spec, assembly out of spec"
Assemblies fail fit or function despite every component meeting its individual tolerances. This recipe guides you through building a tolerance chain, selecting an analysis method (worst case, RSS, Monte Carlo), and redesigning tolerances and datums around real functional requirements.
Create a skill called "Tolerance Stack-Up Detective". Input: (1) assembly functional requirement, (2) dimensioned contributors with tolerances, (3) datum/locating scheme, (4) any known process capability notes. Output: - Identify the correct functional tolerance chain (with sign convention) - Run worst-case and RSS calculations (and Monte Carlo if justified) - Highlight the top 20% contributors causing 80% of the variation - Recommend redesign options: tolerance reallocation, datum changes, adjustability, process changes, or inspection/assembly changes - Provide an inspection plan aligned to the functional chain If any detail is missing, mark it "unspecified" and proceed with bounded assumptions.
This skill converts an assembly requirement (gap, alignment, preload, runout, travel)
into a tolerance chain, then evaluates whether the design is robust to manufacturing
variation. It prioritizes functional dimensioning and early discovery of the kinds of
problems that only show up when you put everything together.
The core issue: no tolerance stack-up was performed on critical functional loops, the
wrong datum scheme was used so the chain doesn't match function, or statistical
assumptions (RSS) were applied without verifying process capability. The result is
prototypes that don't assemble without rework, shimming, or selective assembly — even
though every individual part passes inspection.
End CMM fights and "multiple correct interpretations"
Prints that are syntactically "GD&T-looking" but functionally ambiguous create disputes between design, manufacturing, and inspection. Parts pass measurement yet fail in assembly. This recipe audits a drawing for datum scheme correctness, GD&T completeness, and inspectability — translating function into datum reference frames, tolerance zones, and measurement plans.
Stop receiving 50x50 logos and watermarked stock photos
Clients send tiny JPEGs of their logo cropped from Facebook, watermarked Google Images screenshots, and photos from 2009 flip phones. This recipe checks every asset the client sends, flags what's unusable, and generates a specific re-request telling them exactly what you need.
Great, personalized and innovative brands for sure
A workflow to create a brand for your product or client using Gemini and OpenAI to generate images, with KiloClaw managing the prompts, analyzing results, and iterating until every brand element comes together.
Keep up with what matters, ignore the hype
Set up a lightweight weekly digest around your stack and interests. A nice starter automation because it shows OpenClaw doing recurring research without requiring a huge workflow or lots of context.