---
description: <One sentence — what this skill does AND when to invoke it. Under 200 chars. Used for auto-invocation matching by Claude.>
allowed-tools: Bash, Read, Edit, Write, Glob, Grep
# Uncomment if this skill has side effects (file writes outside CWD, git mutations, network mutations, etc.):
# disable-model-invocation: true
# Uncomment if the skill takes arguments:
# argument-hint: <argument-format>
# Uncomment if research-heavy — runs in forked subagent context with prompt cache inherited:
# context: fork
# agent: Explore
---

# <Skill Name>

**Purpose:** <One sentence describing the capability.>

**When to invoke:**
- Tyler says: "<trigger phrase 1>" / "<trigger phrase 2>"
- <State that auto-triggers this skill, if any>
- <Other invocation conditions>

**Cross-reference:** `~/.claude/rules/slash-command-tool-sequencing.md`, `~/.claude/rules/skill-management.md`.

---

## Phase 1: Load Context (MANDATORY)

**This phase runs before any subsequent phase. No writes, no AskUserQuestion, no commits in this phase.**

### Step 1: Enumerate source files

```bash
# Enumerate all relevant files. NO judgment-skipping — read every file that matches.
find <path> -name "<pattern>" -type f | sort
```

### Step 2: Read all enumerated files

Read every file from Step 1.

**Parallel-call hint (mandatory for >3 files):** Launch the Read calls in a single message so they execute in parallel.

For >15 files OR >3 distinct file types, dispatch parallel subagents via the Agent tool instead — each subagent reads a slice and returns a summary.

### Step 3: Emit status table

After all reads complete, emit a status table:

```
| File | Read | Summary |
|---|---|---|
| <path1> | yes | <one-line summary> |
| <path2> | yes | <one-line summary> |
```

### Pre-write gate (MANDATORY before Phase 2)

Before any subsequent phase that writes, modifies, or invokes AskUserQuestion, confirm:

- [ ] Have I read all source files enumerated in Step 1?
- [ ] Have I checked prior submissions / prior work for this task?
- [ ] Have I loaded the rubric / spec / standard for this task?

yes/yes/yes required before proceeding to Phase 2.

---

## Phase 2: <Workflow phase name>

### Step 1: <Imperative action>

<Specific instructions. Imperative verbs only.>

### Step 2: <Next action>

<Specific instructions.>

---

## Phase 3: <If applicable>

<Continue with numbered Step headers.>

---

## Output

<What the skill produces — file path, report format, etc.>

Example:
```
Dated report at ~/.claude/logs/<name>/<name>-YYYY-MM-DD.md. Printed to stdout.
```

---

## Integration

<Who invokes this skill. What it feeds into. Cross-references to related skills.>

---

## Single-Option Default Reminder

Per `~/.claude/rules/single-option-default.md`: if this skill ever needs to ask Tyler something, give ONE recommendation, never a 3-option menu. Reach for AskUserQuestion only when one of the four ask-triggers fires (see `~/.claude/rules/autonomous-authority.md`).

---

*Authored from `~/.claude/skills/SKILL-TEMPLATE.md` (2026-05-11 version). Conforms to `~/.claude/rules/slash-command-tool-sequencing.md` and `~/.claude/rules/skill-management.md`.*

<!--
Template authoring notes (delete from final skill):

- The `description:` field is what Claude reads to decide when to auto-invoke this skill. Make it specific. "Audit the registry" is bad. "Audit registry for stale agent entries; flag agents in agents.yaml not in CLAUDE.md master table" is good.
- Keep under 500 lines total. If you cross 500, refactor: move reference content to supporting files (e.g., `templates/foo.md`) and Read them on-demand, or split into multiple skills.
- Skills with side effects MUST set `disable-model-invocation: true`. Side-effect categories: file writes outside CWD, git mutations, network mutations, filesystem destructive, external API writes, auth-bearing operations.
- Skills that ONLY read should NOT have `disable-model-invocation: true` — that blocks auto-invocation and is wrong for read-only audits.
- The Phase 1 context-load + status table + pre-write gate is mandatory for any skill touching >1 file. The only exceptions: pure dashboards (read-and-render), pure preprocessing skills with no decisions to make.
- Reference exemplars: `~/.claude/skills/1000xagent-validate-folder/SKILL.md`, `~/.claude/skills/1000xagent-audit-identity/SKILL.md`.
-->