Number every claim so you can strike the wrong ones
It has your agent list every claim in a plan or sentence, the unstated ones first, numbered, so you can strike the wrong ones by number and it rebuilds from what's left.
Free, to use and to change. What it can see: only the chat you add it to.
Its short name: claims. Source on GitHub.
Who it helps
Install
Pick the one you use.
Claude Code
/plugin marketplace add mbsi-labs/skills
/plugin install mbsi@skills
This adds every skill, and each one shows as mbsi:claims. To add only this skill, use this instead, and not both, or the skill shows twice:
/plugin marketplace add mbsi-labs/skills
/plugin install claims@skills
Changelog
- 2026-11-02 (069058e): Updated.
- 2026-10-22 (8cf6617): Updated.
- 2026-10-16 (9d94e6e): Updated.
- 2026-10-16 (e0bb410): Updated.
Licence: MIT.
Read the skill itself
Misalignment hides in implicit claims: things a sentence assumes but never says. Claudes sideload these implicit claims during design, planning, writing, and prioritization tasks, and these implicit claims pollute product sense and cause massive misalignment.
The user can't correct what isn't written, so write all of it, numbered, and let them strike by number to get onto the same page.
- Quote the text verbatim (yours or theirs), so both of you are pointing at the same words.
- List every claim in STE, one idea per line: explicit claims first, then implicit ones. Number them
claim-1…claim-n. Implicit claims matter most. - Count the nouns the user named. Each noun in their request (HTML, sections, interactions) is an explicit claim that the work holds it; check each one has a place. A field set or scope copied from an earlier tool carries that tool's narrower question as an implicit claim.
- Then list APC (axioms, principles, cruxes), with numbering restarting at
apc-1, so a correction can nameclaim-norapc-nwith no ambiguity. - Take the strikes as given. "5, 9, and 11 are incorrect" means rebuild from their negations. Don't defend, don't guess a new belief wholesale: list the claims of your corrected sentence again, and repeat until none is struck.
- Don't ratchet a corrected belief into a rule or skill until the user says it's right. Wording it is part of getting it right.
- No hedges. State a limitation once, where it applies.