Agent skills loaded by Claude Code.
Skills live under skills/, grouped into buckets:
engineering/— daily code workproductivity/— daily non-code workflow toolsmisc/— kept around but rarely usedpersonal/— tied to my own setup, not promotedin-progress/— drafts not yet ready to shipdeprecated/— no longer used
Each skill is its own directory containing a SKILL.md (with YAML frontmatter — name and description) and any bundled scripts or reference files.
npx skills add thananon/9arm-skillsFor my own dev loop — symlink every shippable skill into ~/.claude/skills/:
./scripts/link-skills.shList every SKILL.md in the repo:
./scripts/list-skills.sh- debug-mantra — Four-mantra debugging discipline: reproduce → trace the fail path → falsify the hypothesis → cross-reference every breadcrumb. Recites verbatim at session start, then applies in order before any fix.
- post-mortem — Write the canonical engineering record of a fixed bug — root cause, mechanism, fix, validation, how it slipped through. Engineer-audience; refuses to draft without a reliable repro, known cause, and validated fix.
- scrutinize — Outsider-perspective end-to-end review of a plan, PR, or code change. Questions intent (is there a simpler way?), traces the actual code path, and verifies the change does what it claims. Output is concise, actionable, with rationale.
- qwen-agent — Delegate menial, well-scoped coding tasks to a cheap Qwen-backed subagent via the
claude-9armcommand. Use for bulk renames, formatting, boilerplate, grep-style summarization, scaffolding, and build/test reporting.
- management-talk — Rewrite engineer-to-engineer content for engineering-org leadership and shape it for the channel it's going to (JIRA, Slack, async standup, email, meeting talking-points).
- qwenchance — Keeps a long Claude Code task on-track — breaks loops, bounds internal reasoning, watches the context budget, and triggers a clean handoff before the window fills.
(none yet)