Skill Issue

SkillAI & models

Triggers when an agent authors, reviews, names, or decides whether to create an agent skill, including requests to write SKILL.md, choose a skill type, evaluate skill-worthy content, or split material into references/scripts. Not for running skill evaluations or description optimization, which belong to skill-creator.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the Skill Issue skill

What this skill tells your AI

The instructions your AI receives, as published by eliasstravik/skills in skills/skill-issue/SKILL.md and read by ahel’s review.

Trigger

Apply this library standard when authoring, reviewing, naming, or deciding admission for an agent skill.

Scope

This SOP governs skill form in this library; skill-creator owns evaluation, iteration, and description optimization. Existing skills predate this standard and stay unchanged unless the client explicitly places them in scope.

Inputs

The form decision accepts the proposed skill's purpose and candidate content; evaluation evidence belongs to skill-creator.

Roles

The authoring agent applies this standard; the client decides admission disputes and taste calls.

Procedure

  1. Reject the skill unless its candidate content contains at least one survivor category in the litmus.
  2. Classify the admitted skill as exactly one of the five types in the typology.
  3. Name the skill by its type's rule.
  4. Write all nine sections in this order: Trigger, Scope, Inputs, Roles, Procedure, Outputs, Exceptions, QC, References; write None. in an inapplicable section.
  5. Reduce the skill to: “When [trigger] occurs, [role] performs [procedure] using [inputs] to produce [output], while satisfying [controls].”
  6. Delete every body line an agent with perfect reasoning and perfect public-world knowledge could derive.
  7. Move bulky non-derivable material to references/, deterministic operations to scripts/, and write the frontmatter description by the description canon.

Outputs

Record compliance as one conforming skill directory containing SKILL.md and only required attachments; this Policy SOP produces no domain work product.

Exceptions

A failed admission gate yields a rejection, not a thinner skill. Escalate admission disputes to the client; never waive the gate silently.

QC

  • Confirm a survivor category earned admission and every retained body line passes the litmus.
  • Confirm the type is singular, the name follows its type, and all nine ordered sections are present.
  • Confirm the name matches its directory, uses lowercase hyphens, and is at most 64 characters.
  • Confirm the body is under 500 lines and 5,000 tokens; each reference over 100 lines has a table of contents.
  • Confirm the non-empty third-person description is at most 1,024 characters, contains no XML, and states positive and negative triggers.
  • Confirm the trigger-eval optimizer's final description is applied verbatim.

References

Read SOP types for classification or naming, the litmus for admission or line review, and the description canon when writing frontmatter.

Signals

GitHub stars
20
Last commit
Aug 2026
Advanced
Catalog kind
skill
Gateway key
skill-issue
Source
github.com/eliasstravik/skills