CtrlK
BlogDocsLog inGet started
Tessl Logo

ideate

The idea field is empty, narrowed too early, or thin in places: widen it in width and depth before any candidate is chosen; reads the user's words and what they name.

51

Quality

64%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./heuresis/skills/ideate/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

48%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The skill is a carefully engineered, internally consistent contract with a genuinely clear phase/gate workflow and well-wired, conditionally-loaded reference files. Its costs are heavy: a monolithic inline formal specification whose content and rules duplicate the prose, leaving the actual per-turn executable guidance thinner and more inferential than the token budget suggests.

Suggestions

Move the Lean formal contract to references/ (e.g. references/contract.lean) and keep in SKILL.md only the operational protocol, rules, and a pointer to the contract, reading it only when a question of the contract's semantics arises — this would address both the conciseness and progressive-disclosure gaps at once.

Deduplicate the Rules section against the Lean doc-comments (frame-first, person-closes, out-of-scope-never-erased, chain semantics all appear in both) by keeping each invariant in exactly one place and cross-referencing.

Add one short worked example of a round presentation (a rendered map with a gate) so the executor has a concrete template rather than synthesizing form purely from the formal spec.

DimensionReasoningScore

Conciseness

The body is ~485 lines, dominated by a 430-line formal Lean 4 contract whose grounding machinery (Turn, Cite, Coord, Occ, lift, etc.) and doc-comments substantially duplicate what the Mode Activation/Protocol/Rules prose restates (frame-first, person-closes, out-of-scope-never-erased, chain semantics all appear twice). This matches anchor 2 ('noticeably verbose; several unnecessary explanations or padded sections') rather than 3 — the duplication and sheer formal apparatus go beyond 'could be tightened' — and not 1 because nothing explains concepts Claude already knows and every section does carry contract content.

2 / 5

Actionability

The Protocol section gives conditionally-timed, concrete directives ('read references/blank-entry.md before presenting abstract frames', 'continue first and stop second', 'one map every round with this round's additions marked') and the U-BINDING block has concrete invocation examples ('/ideate "topic or fragments" → U = the argument'). But the operational behavior itself is spread across a formal, partly-axiomatized spec that the executor must synthesize into per-turn actions, and no worked example transcript shows what a round's output actually looks like — matching anchor 3 ('some concrete guidance but incomplete... missing key details') rather than 4, where guidance would be mostly directly executable.

3 / 5

Workflow Clarity

The FLOW comment, phase transitions, and Protocol section give a clear, ordered sequence: relay/classify → Blank path (frame map, gate) vs Seeded path (first pass) → per-round fuse/reference-read/closing-check/targets/pass/present/gate → Stop/withdraw/route terminals, with explicit checkpoints (classification relay with basis, gates before each round, dissent and unaddressed signals surfaced before every gate, EarlyExit fully declared). This matches anchor 4 ('clear sequence with most checkpoints present; minor validation gaps') rather than 5 because no error-recovery feedback loop is specified (e.g. what to do when a relay's basis or a map's density is judged wrong), and rather than 3 because checkpoints are explicit, not implicit.

4 / 5

Progressive Disclosure

All four bundle files (blank-entry.md, seeded-entry.md, chain-reference.md, round-composition.md) are real, one level deep (no nested references), and clearly signaled with precise read-conditions ('read X before presenting Y'). However, the bulk of the skill — the ~430-line Lean contract — is inlined in SKILL.md when it could live in a reference file loaded on demand, matching anchor 3 ('some structure but... content that should be separate is inline') rather than 4, which would require most content appropriately split across files.

3 / 5

Total

12

/

20

Passed

Description

62%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description correctly conveys a niche capability — divergent widening of an underexpanded idea field before selection — with an explicit when-condition, and it is free of fluff. Its main weaknesses are thin concrete action coverage and the absence of the natural trigger vocabulary (brainstorm, ideas, options, alternatives) users would actually say.

Suggestions

Add an explicit 'Use when...' clause with natural trigger phrases users say, e.g. 'Use when the user asks for ideas, wants to brainstorm, says they are stuck, or feels the options so far are too narrow.'

Name 2-3 concrete outputs of the widening (e.g. alternative frames, deepened branches under open candidates, a live map of the field) so the 'what' covers what the person actually receives.

Include common synonyms — brainstorm, ideas, options, alternatives — as trigger terms to improve keyword coverage for skill matching.

DimensionReasoningScore

Specificity

The description names the domain (an 'idea field') and one concrete operation — 'widen it in width and depth before any candidate is chosen' — but stops at 1-2 actions and never enumerates what the widening actually produces (frames, branches, candidates, a map). It sits at anchor 3: a clear what-ish action but not comprehensive coverage; not anchor 2 because the widening action and the reads-the-user's-words qualifier are more specific than a generic 'processes ideas', and not anchor 4 because no several-action list is present.

3 / 5

Completeness

The 'what' is present ('widen it in width and depth before any candidate is chosen') and the 'when' is explicit as a front-loaded condition ('The idea field is empty, narrowed too early, or thin in places'), so both halves are answered. It fits anchor 4 ('both what and when; when could be more explicit') rather than 5 because there is no 'Use when...' trigger-phrase formulation tied to user utterances, and rather than 3 because the when-clause names three concrete precondition states rather than being missing or merely implied.

4 / 5

Trigger Term Quality

Terms like 'idea field is empty', 'narrowed too early', 'thin in places', and 'widen' are words a user might plausibly use when stuck for ideas, so some natural keywords exist. But common variations users actually say — 'brainstorm', 'give me ideas', 'more options', 'alternatives' — are absent, matching anchor 3 ('some relevant keywords but missing common variations or synonyms').

3 / 5

Distinctiveness Conflict Risk

The niche — divergent widening before any selection, explicitly 'before any candidate is chosen' — is distinct from evaluation/decision/research skills and the phrasing is idiosyncratic enough to avoid accidental triggering. This matches anchor 4 (mostly distinct, minor overlap risk with closely related skills such as general brainstorming or planning skills) rather than 5, because no terms sharply fence it off from adjacent 'help me think of options' skills.

4 / 5

Total

14

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
jongwony/epistemic-protocols
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.