Skip to content

feat(setup-notifications): default to Notification-only, make Stop opt-in#1

Closed
craftengineer wants to merge 207 commits into
mainfrom
feat/notifications-default-quiet
Closed

feat(setup-notifications): default to Notification-only, make Stop opt-in#1
craftengineer wants to merge 207 commits into
mainfrom
feat/notifications-default-quiet

Conversation

@craftengineer

Copy link
Copy Markdown
Owner

Why

The Stop hook fires at the end of every assistant turn — Claude Code has no multi-turn "task done" concept, so a chatty back-and-forth session pings after each reply even with throttling. That's the noise source.

The Notification event already covers both permission prompts and ~60s of true idle, so on its own it delivers the "come back, I need you" ping — which is what most users actually want.

What changed

skills/setup-notifications/SKILL.md:

  • Recommended default is now Notification only; Stop is documented as an opt-in per-turn ping with an explicit note that it fires at every turn end.
  • Updated: frontmatter description, intro, the events question, the register-hooks block (Notification always added, Stop optional), the verify steps, and the notes/rollback.

No change to assets/notify-webhook.sh — the script already handles both events; only the default wiring/recommendation changes.

🤖 Generated with Claude Code

jsifalda and others added 30 commits March 8, 2026 14:56
Interactive keyword strategy generator for side projects.
Walks through project analysis (value prop, audience, emotional drivers)
then generates keywords in 8 categories with top-10 prioritization.
Extracts atomic tasks from Obsidian notes and appends them to To Remember.md
with proper formatting, due dates, and 8-week recurring intervals.

- SKILL.md with full extraction process and task format spec
- scripts/extract_tasks.py helper for task formatting
5-advisor council system for pressure-testing decisions:
- Contrarian, First Principles, Expansionist, Outsider, Executor
- Anonymous peer review round
- Chairman synthesis with final verdict
- HTML report + markdown transcript output
- Triggers: 'council this', 'war room this', 'pressure-test this'
feat: add council skill (Karpathy LLM Council method)
Convert the always-apply educator rule into an on-demand skill with
structured workflow, STARTHERE.md template, and auto-detection of
the project's agent instruction file (CLAUDE.md, AGENTS.md, .cursorrules, etc).

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Co-authored-by: Copilot <[email protected]>
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
jsifalda and others added 27 commits July 3, 2026 14:27
…-finder

refactor: rename youtube-video-finder skill to yt-video-finder
Adopt the golden-template structure (problem-first, metrics with baseline/guardrails, scope in/out, risks, rollout), target 1-2 pages, and save PRDs to docs/prds/ linked from the project README. Keeps the junior-dev implementable framing.
Drop rollout, evidence, and corporate metrics/dependency ceremony; reframe for planning your own side-hustle features. Keeps buildable requirements, docs/prds save path, and README linking.
…t-practices

refactor(prd-creator): realign PRD template to best practices
- Adds a PRD-first gate: require a PRD via /prd-creator before substantial features, draft the plan from it, and read docs/prds/ for context before any plan.
- Opt-in per project, gated to substantial features only.
Installs Claude Code Stop/Notification hooks that POST to a webhook (e.g. n8n),
so you get pinged when a session finishes a task or is waiting on your input —
useful when running Claude Code in tmux over mobile SSH.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
…te-like-human

refactor(indie-hacker-wrapup): load write-like-human skill before drafting
…-prd-gate

feat(setup-aiengineering): add opt-in PRD gate module
…fications-skill

feat: add setup-notifications skill
Indie builders should skim a PRD fast. Added a conciseness contract,
per-section budgets, bullet-point-default format, and a worked example;
banned corporate filler. Kept all 8 sections so downstream skills still parse.
After a PRD is saved, offer to stress-test it with grill-me, iterate,
then fold resolved decisions back in (confirmed before writing).
Reference is guarded, skipped silently when grill-me is absent, so the
skill stays portable.
…ser-prds

feat(prd-creator): terser PRDs + optional grill-me pressure-test
…-docs-alignment-gate

feat(setup-aiengineering): add docs-alignment gate to verification protocol
- Converts text-based PDFs to clean structured Markdown without OCR: layout-aware extraction, page-furniture stripping, paragraph reflow, structure-to-heading mapping.
…rify-loop

feat: add bounded verification loop to write-like-human skill
…s-1783712445399

Add Claude Code GitHub Workflow
…t-in

The Stop hook fires at the end of every assistant turn (Claude Code has no
multi-turn "task done" concept), so a chatty session pings after each reply
even with throttling. Notification already covers both permission prompts and
~60s true idle, so it alone delivers the "come back, I need you" ping.

Make Notification the recommended default and document Stop as an opt-in
per-turn ping. Updates the description, intro, events question, register-hooks
block, verify steps, and notes.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
@craftengineer

Copy link
Copy Markdown
Owner Author

Closing — this should live on jsifalda/ai-project-rules instead.

@craftengineer craftengineer deleted the feat/notifications-default-quiet branch July 11, 2026 09:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants