SkillAgentSearch skills...

taste

Use when asking "overkill?", "elegant?", "audit", or "taste-test this", or setting a taste register. Not for tasks that require source or remote-system changes.

Install / Use

npx skills add OutlineDriven/outline-driven-development --skill taste

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

65/100

Supported Platforms

Universal

Tags

Our assessment of taste

taste scores 65/100 on our quality scale, 3976th of 4,599 Development & Engineering skills we index.

Its SKILL.md is 6.3 KB long, well organised into 14 sections with 1 code example: a thorough specification that gives an agent plenty to work with.

It has no GitHub stars yet, so there is no community track record; judge it on its content.

Substance
29/30
Structure
17/20
Description
15/15
Adoption
0/20
Freshness
5/15

Maintenance, license and trust

  • We could not determine when the repository was last updated.
  • Our last check on 2026-09-04 found the source still online.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 68/100, with 3 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

taste compared with similar skills

All 4 of these similar skills score higher than taste; compare them before choosing.

SkillScoreStarsUpdatedFormat
taste (this skill)by OutlineDriven650—SKILL.md
taste-skillby Leonxlnx100102.7ktodayCLAUDE.md
ai-job-searchby ai-job-search-dev10066.0k5d agoCLAUDE.md
claude-howtoby luongnv8910045.2k6d agoCLAUDE.md
algorithmic-artby anthropics100177.9k34d agoSKILL.md

Frequently asked questions

How do I install taste?
Run npx skills add OutlineDriven/outline-driven-development --skill taste. The install tabs above show the steps for each supported agent.
Which AI agents does taste work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is taste safe to use?
It declares no license and scores 68/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is taste still maintained?
We could not determine when the repository was last updated.

name: taste description: 'Use when asking "overkill?", "elegant?", "audit", or "taste-test this", or setting a taste register. Returns a per-anchor verdict table with ranked fixes in audit mode, or loads the judgment register in anchor mode. Don''t use for tasks that require source or remote-system changes.'

Taste

Contract

| Field | Bound contract | |---|---| | Trigger | User asks 'overkill?', 'elegant?', 'audit', or 'taste-test this', or asks to set a taste register | | Authority | Read-only. No file, VCS, credential, paid, published, deployed, or remote mutation. | | Side effect | Audit returns a verdict table and ranked fixes in chat; anchor mode persists the register best-effort; no files are written. | | Done | A per-anchor verdict table with ranked top-3 fixes is returned, or the taste register is loaded in anchor mode. |

Inputs

  • The artifact or decision to judge (required for audit mode).
  • No inputs required for anchor mode beyond the invocation.

Procedure

Mode selection

Auto-detect from the user's phrasing, with slash-arg override:

  1. User wording matches overkill?, elegant?, audit, taste-test this, judge this → audit mode.
  2. User wording matches taste anchor, taste mode, taste register, or anticipates producing fresh work → anchor mode.
  3. Anything else → audit (default; cheaper to run; no behavior commitment).
  4. Explicit override: /taste audit or /taste anchor. Override always wins.

Audit mode

  1. Walk the eleven anchors one at a time against the artifact.
  2. For each anchor: state the anchor, judge the artifact (pass / warn / fail), cite the Side A or Side B charter row when violated, and write a concrete fix.
  3. When two anchors fail with conflicting fixes (e.g., Restraint says compress, Generosity says expand), surface the tension explicitly. Do not auto-pick. Tie-break is user-led.
  4. Close with the top-3 ranked fixes.

Anchor mode

  1. Load the charter and anchors as imperatives for subsequent responses.
  2. Persistence is best-effort: applies until the user signals "stop taste" or "normal mode" OR context is compacted, whichever comes first.
  3. Re-invoke /taste anchor if drift is observed.

Auto-clarity exception

Suspend the register temporarily for:

  • Destructive or irreversible operation confirmations (e.g., git push --force, rm -rf).
  • Security or data-loss warnings.
  • Multi-step procedures where order or atomicity matters and the judgment register would obscure structure.
  • Direct user clarification requests.

Resume the register once the high-stakes section ends.

The two-sided charter

Side A — slop (centroid-AI default convergence): generic openers ("Sure!", "Of course"), hedge-stacks ("perhaps it might be"), validation phrases ("you're absolutely right"), AI-flat prose with no rhythm, default palettes, defensive nil-checks where impossible, 50/50 decision hedges that pick nothing.

Side B — overkill (decoration covering thin ideas): gradient stacks on every section, thesaurus-soup prose ("orchestrate the comprehensive synthesis of"), abstraction towers (4 layers where 1 suffices), complexity-flex masking absent conviction, ceremony that performs depth without delivering it.

| Domain | Side A (slop) | Side B (overkill) | |----------|------------------------------------------|------------------------------------------| | Prose | "I think this might possibly help..." | "We orchestrate a paradigm shift across" | | Code | try { x } catch { /* swallow */ } | Factory<Builder<Strategy<T>>> | | Design | Default purple-blue gradient | Gradient on every section + glow + glass | | Decision | "Both options have merit, so..." | 12-criterion weighted scoring matrix |

Anchors

Eleven anchors apply across prose, code, design, and decisions:

| Anchor | Directive | |---|---| | Clarity | The artifact says what it means; reader does not have to decode. | | Hierarchy | Important looks important; secondary supports. | | Intent | Every choice is committed; nothing reads as "I let the default decide." | | Coherence | Parts agree; tension only where deliberately staged. | | Restraint | Default posture; compress before adding. | | Generosity | Gives more than required at the right moment. | | Honesty | No decoration covering missing depth; no slop covering missing POV. Care finishes the unseen. | | One strong moment | Exactly one commitment carries the lift; the rest supports. | | Invariants | The artifact names what must hold; preconditions, postconditions, and invariants are explicit. | | Audience | The artifact enters through the reader's frame, not the author's. | | Emptiness | The artifact leaves space the user completes; the least assertion. |

Failure and recovery

| Failure class | Response | |---|---| | Artifact too short or ambiguous to judge | Return "insufficient surface" with the specific missing element; do not fabricate verdicts. | | Anchors produce conflicting fixes | Surface the tension; do not auto-pick. Tie-break is user-led. | | Anchor mode drift after context compaction | Re-invoke /taste anchor; do not assume persistence. | | User artifact is in a domain with no charter row | Apply the anchors directly; note the domain gap in the verdict table. |

No partial-result fabrication. If fewer than eleven anchors can be judged, return only the judged anchors and name the blocker.

Output

Audit mode

Anchor             | Verdict | Citation              | Fix
-------------------|---------|-----------------------|----------------------------
Clarity            | pass    |                       |
Hierarchy          | warn    | Side A: AI-flat prose | Lead with the verdict line
Intent             | fail    | Side A: hedge-stack   | Pick one; drop "might"
Coherence          | pass    |                       |
Restraint          | warn    | Side B: ceremony      | Cut the framing paragraph
Generosity         | pass    |                       |
Honesty            | pass    |                       |
One-strong-moment  | fail    | Side A: 50/50 hedge   | Commit to one direction

Top-3 fixes: 1. Pick one direction (Intent + One-strong-moment).
             2. Lead with the verdict (Hierarchy).
             3. Cut the framing paragraph (Restraint).

Anchor mode

Related Skills

View on GitHub
GitHub Stars0
CategoryDevelopment
UpdatedNaNy ago
Forks0

Trust signals

68/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

2 medium1 low