Menu
Surface unnamed patterns across vault notes. Use when: (1) /emerge for last 30 days, (2) /emerge 14d for custom window, (3) /emerge this week.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Use when the user wants a thorough, high-assurance review of code changes — phrases like "review this until it's clean", "converge to zero issues", "adversarial review", "have Gemini and Claude review", "deep review this PR", or "make this change ironclad". Runs TWO phases on a PR or working-tree diff: (1) iterative multi-reviewer review that loops fix->re-review until a round finds zero actionable issues, then (2) a multi-round Gemini-primary adversarial cross-examination (Gemini finds -> Claude judges -> Gemini counters), fixing every confirmed finding. Repeatable across any project/PR. Use when: (1) the user wants a thorough, high-assurance review that converges to zero actionable issues, (2) the user asks for an adversarial or Gemini-and-Claude cross-examination review of a code diff, (3) deep-reviewing a PR or working-tree diff before merge, (4) the user wants to make a change ironclad.
Publishes Claude Code skills as installable plugins and syncs them to a GitHub monorepo. Plugin-first: every skill with a plugin-manifest.json is auto-assembled and synced as a plugin. Also supports bare skill publishing and individual repos. Use when: (1) user says 'publish', 'share', or 'sync' a skill, (2) a skill needs to be made installable by others, (3) syncing skills/plugins to the monorepo, (4) creating a versioned monorepo release, (5) assembling a plugin from skills + commands, (6) user says 'publish plugin' or 'package plugin'.
Creates detailed story specification files from various inputs (Claude plan, requirement file, prompt, GitHub issue). Discovers project spec conventions at runtime, brainstorms approaches with vertical splitting recommendations for large stories, generates template-compliant specs, checks for over-engineering, and optionally chains to spec-review. Use when: (1) user wants to write a new story spec, (2) converting a plan or requirements into a formal spec, (3) creating specs from GitHub issues, (4) breaking a large feature into shippable vertical slices.
Reviews and enriches story specifications with codebase-verified technical sub-tasks, architecture alignment checks, design simplification suggestions, and API test plans. Dynamically discovers project architecture at runtime. Use when: (1) a new story spec needs review before implementation, (2) a spec has high-level tasks but lacks implementation-ready detail, (3) need to verify spec assumptions against actual codebase, (4) a spec references API changes but has no test plan, (5) reviewing specs that reference data shapes or pipeline ordering, (6) spec subtasks mention add field X to object Y or call function at line N.
Creates detailed story specification files from various inputs (Claude plan, requirement file, prompt, GitHub issue). Discovers project spec conventions at runtime, brainstorms approaches with vertical splitting recommendations for large stories, generates template-compliant specs, checks for over-engineering, and optionally chains to spec-review. Use when: (1) user wants to write a new story spec, (2) converting a plan or requirements into a formal spec, (3) creating specs from GitHub issues, (4) breaking a large feature into shippable vertical slices.
Implements a previously created and reviewed story spec end-to-end: reads the spec, creates a feature branch, implements all sub-tasks with progress tracking, validates acceptance criteria, updates tracking files, and creates a PR. Optionally delegates to separately-installed brainstorming, frontend-design, and ui-from-requirements skills for complex UI work. Use when: (1) user says /spec-implement or 'implement this spec', (2) a story spec has been created and reviewed and is ready for implementation, (3) user provides a spec file path to implement.
| name | emerge |
| description | Surface unnamed patterns across vault notes. Use when: (1) /emerge for last 30 days, (2) /emerge 14d for custom window, (3) /emerge this week. |
| metadata | {"version":"1.0.0"} |
Tools needed: Bash, Agent, Write, Read
TaskCreate: subject="Collect and upgrade vault corpus", activeForm="Collecting vault corpus"
TaskCreate: subject="Analyze patterns across notes", activeForm="Analyzing patterns"
TaskCreate: subject="Build emerge report", activeForm="Building report"
TaskCreate: subject="Write vault note", activeForm="Writing vault note"
TaskCreate: subject="Present results", activeForm="Presenting results"
Track task IDs. Set task #1 to in_progress.
Optional flags:
--include-snapshots — include claude-snapshot notes in the corpus.
By default snapshots are excluded; their transient "Key context" bullets
dilute cross-session pattern synthesis.Parse args: strip --include-snapshots first (set INCLUDE_SNAPSHOTS=1 or 0), then parse remaining arg as DAYS: no arg=30, Nd/N days=N, this week=days since Monday.
cd "$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
python3 -c '
import sys, os, glob; sys.path.insert(0, max(glob.glob(os.path.expanduser("~/.claude/plugins/cache/*/obsidian-brain/*/hooks")), default="hooks"))
from emerge_cli import run_corpus; run_corpus(int(sys.argv[1]) if len(sys.argv) > 1 else 30, include_snapshots=(sys.argv[2] == "1") if len(sys.argv) > 2 else False)
' "$DAYS" "$INCLUDE_SNAPSHOTS"
If STATUS starts with CACHED:, report "Using cached corpus (< 15 min old, same window)" and skip to Step 2.
Parse STATUS (OK:<total>:<upgraded>:<failed> or EMPTY:0:0:0). EMPTY -> tell user to widen window, stop. Report upgrades. If failed > 0 -> Step 1f, else Step 2. Mark task #1 completed.
Spawn parallel sub-agents for failed notes (read note, write summary to temp, write back via upgrade_note_with_summary) — mirrors /recall Phase 2 fallback. Re-collect:
cd "$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
python3 -c '
import sys, os, glob; sys.path.insert(0, max(glob.glob(os.path.expanduser("~/.claude/plugins/cache/*/obsidian-brain/*/hooks")), default="hooks"))
from emerge_cli import run_recollect; run_recollect(int(sys.argv[1]) if len(sys.argv) > 1 else 30, include_snapshots=(sys.argv[2] == "1") if len(sys.argv) > 2 else False)
' "$DAYS" "$INCLUDE_SNAPSHOTS"
Set task #2 to in_progress. Spawn one Agent:
Agent({
description: "Analyze vault corpus for cross-cutting patterns",
prompt: "Read ~/.claude/obsidian-brain/emerge-corpus.json. Analyze and write to ~/.claude/obsidian-brain/emerge-analysis.md with these 5 categories:\n\n## Recurring Technical Patterns\nPatterns in approaches, tools, or architecture across 2+ sessions.\n\n## Error Clusters\nRepeated error types, common root causes, or failure modes.\n\n## Decision Trends\nDirectional shifts in technical decisions over time.\n\n## Cross-Project Connections\nShared themes between projects. Skip if only 1 project.\n\n## Emergent Practices\nImplicit conventions that formed organically but aren't documented.\n\nFor each pattern: descriptive name, 2-3 examples with note references, confidence (strong/moderate/tentative).\n\nWrite using the Write tool. Return ONLY: WRITTEN:~/.claude/obsidian-brain/emerge-analysis.md"
})
If no WRITTEN: response, report failure and stop. Mark task #2 completed.
cd "$(git rev-parse --show-toplevel 2>/dev/null || pwd)"
python3 -c '
import sys, os, glob; sys.path.insert(0, max(glob.glob(os.path.expanduser("~/.claude/plugins/cache/*/obsidian-brain/*/hooks")), default="hooks"))
from emerge_cli import run_build_note; run_build_note()
' 2>&1
Parse SAVED:<path> and everything after ---REPORT---. Mark tasks #3-#4 completed.
Display report prefixed with Pattern Discovery Results:. Confirm saved path. Mark task #5 completed.
/emerge 60d)./obsidian-setup first.