From thinking-frameworks-skills
Classifies hedges in drafts as precision (keep) or weakness (flag), suggests rewrites. Use when a draft feels wishy-washy.
How this skill is triggered — by the user, by Claude, or both
Slash command
/thinking-frameworks-skills:hedge-detectorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
- [Precision vs weakness](#precision-vs-weakness)
Related skills: Called by the Editor in the voice pass. Complements voice-check (which flags "I think" as a don't-list phrase when used as primary hedge). This skill does the finer classification.
Precision hedge (KEEP): scope-naming, sample-size-caveat, specific-uncertainty.
Epistemic-weakness hedge (FLAG): softens without adding information.
For each hedge in the draft:
- [ ] Step 1: Detect hedge markers (modal verbs + phrase list above)
- [ ] Step 2: Classify as precision or weakness
- [ ] Step 3: For weakness, suggest a commit OR a specific hedge (both, as 2 rewrite options)
- [ ] Step 4: For precision, leave alone (note in the "calibrated hedges kept" count)
- [ ] Step 5: Emit the hedge audit with both lists
A hedge is precision if paired with specific bounds:
Otherwise weakness. Default to weakness when unsure — the writer prefers over-flagging here.
For each weakness hedge:
Both options; writer picks.
Draft sentences:
Classification:
| # | Hedge | Class | Rewrites |
|---|---|---|---|
| 1 | "I think" | weakness | (a) "RAG beats fine-tuning for most teams." (b) "In the three teams I've worked with, RAG beat fine-tuning." |
| 2 | "I do not know" + scope | precision | Keep as-is. |
| 3 | "Arguably" | weakness | (a) "The attention mask is wrong." (b) "The attention mask looks wrong to me — I have not re-derived the gradient." |
| 4 | "Perhaps" + "very specific" | weakness | (a) "Fine-tuning wins on style." (b) "Fine-tuning wins on style; I have not tested this below 7B." |
slop-detector signal S8.slop-detector S8.npx claudepluginhub lyndonkl/claude --plugin thinking-frameworks-skillsScans a draft for 10 signatures of AI-generated explainer slop including meta-framing openers, zombie nouns, prompt residue, and hedge clusters. Use when a draft feels generic.
Audits and rewrites content to remove AI writing patterns ("AI-isms"). Supports detect-only mode, file edit-in-place, voice profiles, and iterative refinement.
Humanizes AI-generated text by detecting and rewriting patterns like inflated symbolism, em dash overuse, passive voice, rule of three, and filler phrases. Use for editing or reviewing docs and code comments.