qhelp

name	Qhelp
description	Shows QE Framework usage overview. With no arg, prints the full catalog. With a skill name arg (e.g., /Qhelp Qcommit), reads that skill's SKILL.md and summarizes it in the user's language. Also supports find/search skills mode via reference/find-skills.md.
invocation_trigger	User asks for help, uses /Qhelp, or invokes any QE skill with --help / -h flag.
recommendedModel	haiku

Qhelp — QE Framework Reference & Per-Skill Help

Role

Three modes:

1. No argument: print the full QE Framework reference card (all skills).
2. With skill argument: read that skill's SKILL.md and generate a per-skill usage summary in the user's language.
3. Find/search skills: when the user asks to find, search, browse, or install skills from skills.sh, follow reference/find-skills.md.

Workflow

Mode A: No argument

Print the full reference card below directly to the user.

Mode B: Skill argument (/Qhelp {skillName})

1. Read .qe/profile/language.md to detect user language. Parse Primary language: <code>. Default to en.
2. Resolve target: skills/{skillName}/SKILL.md.
3. If not found: glob skills/*/SKILL.md, compute 3 closest matches by name similarity, output not-found message + suggestions in user's language. Stop.
4. Read the target SKILL.md. Extract:
   • frontmatter description → Section 1
   • frontmatter invocation_trigger (or fall back to description) → Section 2
   • ## Role section + first workflow/behavior block → Section 3 (3–5 bullets)
   • code blocks containing /{skillName} → Section 4 (or default /{skillName} if none)
5. Output 4 sections with headings in the user's language:
   • ko: "한 줄 요약" / "언제 쓰나" / "주요 동작" / "사용 예시"
   • en: "Summary" / "When to use" / "What it does" / "Usage"
   • ja: "概要" / "使用タイミング" / "主な動作" / "使用例"
   • other: English
6. Keep total output ≤400 words.

Mode C: Find/search skills

When the request is about finding a skill, searching skills.sh, browsing available skills, or installing a discovered skill, read and follow reference/find-skills.md. This mode searches skills.sh, fetches original SKILL.md content, converts it to a QE-compatible SKILL.md, confirms the install location through the QE interaction adapter, writes the file, and verifies installation.

Reference Card (Mode A output)

QE Framework (Query Executor) — Quick Reference
================================================

WORKFLOW
  /Qgenerate-spec          Create task spec documents
  /Qexecute [UUID]        Execute tasks with verification ($Qexecute on Codex)
  /Qcommit                 Commit & push (human-style, no AI traces)
  /Qupdate                 Update QE (Claude + Codex) and codex-plugin-cc bridge
  /Qexecute -utopia [on|off]        Toggle fully autonomous mode ($Qexecute -utopia on Codex)

DEVELOPMENT
  /Qexecute -verify          Quality loop: test → review → fix ($Qexecute -verify on Codex)
  /Qmcp setup              MCP server setup + building guide
  /Qmcp ensure             Check MCP config, server availability, stale entries

TASK MANAGEMENT
  /Qinit                   Initialize QE in a project
  /Qcollect-skill          Collect verified local stack skills into .claude/skills/
  /Qrefresh                Refresh project analysis
  /Qresume                 Restore previous session context
  /Qcompact                Save context / session handoff ($Qcompact on Codex)
  /Quser-action            Track external actions the user must perform
  /Qgc archive             Archive completed tasks

PLANNING
  /Qplan                   Plan a milestone/project
  /Qgs                     Generate executable task specs
  /Qqa-test-planner        QA test plans & bug reports

META
  /Qversion                Show current version
  /Qhelp find              Search skills.sh marketplace
  /Qalias                  Path & command aliases
  /Qhelp                   This help screen

ADMIN
  qe-admin-mcp             Maintainer-only workflows
                           (qe_admin_search_skills / qe_admin_read_skill)

AGENTS (auto-selected by complexity)
  HIGH   Edeep-researcher, Eqa-orchestrator
  MEDIUM Etask-executor, Ecode-debugger, Ecode-reviewer,
         Ecode-test-engineer, Edoc-writer,
         Egrad-writer, Epm-planner, Erefresh-executor,
         Ecompact-executor, Ehandoff-executor
  LOW    Earchive-executor, Ecommit-executor

Will

• Display the quick reference card (Mode A)
• Read a specific skill's SKILL.md and generate a 4-section summary in the user's language (Mode B)
• Search skills.sh and install discovered QE-compatible skills by following reference/find-skills.md (Mode C)
• Provide skill-name suggestions when a requested skill is not found

Will Not

• Execute any commands from the target skill
• Modify files in Modes A or B
• Translate the reference card itself (Mode A stays in its native mixed form)

Skill Budget Status

When displaying help, include a budget summary line at the top:

Skills: {count} loaded | Budget: ~{tokens} tokens ({pct}% of 1M context)

This information is calculated by hooks/scripts/lib/skill-budget.mjs:

• calculateSkillBudget(skillsDir) — counts SKILL.md files and estimates token usage
• checkBudgetOverflow(totalTokens) — checks if budget exceeds 1% of context window
• getCompressibleSkills(skillsDir) — lists skills with longest descriptions for compression

If budget overflows (>1% of context), display a warning:

⚠ Budget overflow — consider running skill deduplication audit