muratsilahtaroglu/claude-code-starter-kit
Agent Claude Code
Subagents are isolated assistants for side tasks (deep research, audits, verification) that would clutter the main thread with intermediate results you won't reference again. Each runs in its own context window and returns only its final message to the parent — so the raw exploration never enters your main context.…
muratsilahtaroglu/claude-code-starter-kit
Agent Claude Code
Read-only rules-compliance auditor — samples a commit range against rules.md (layout §3, security §5/§7, tests §2, docs-sync §1, memory §9) and returns violations with file:line + the rule §, severity-ranked. Reports only, never edits. Spawned by /keel-audit; the SessionStart hook nudges when one is due.
muratsilahtaroglu/claude-code-starter-kit
Agent Claude Code
Isolated external-research agent — scans the web and the local repo for prior art, returns a distilled, CITED findings summary (not raw pages). Use when a question needs docs/papers/GitHub/community input that would clutter the main thread. Backs rules.md §8 and the /keel-research skill.
muratsilahtaroglu/claude-code-starter-kit
Agent Claude Code
Adversarial verification agent — given a claim, finding, or proposed fix, tries to REFUTE it (reproduce the failure, run the test, read the code) and returns CONFIRMED / REFUTED / UNCERTAIN with evidence. Use to check sub-agent output, review findings, or "is this actually true?" before acting. Backs rules.md §4 and…
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs before the agent uses a tool for Bash, Skill, Edit, Write, NotebookEdit and SendMessage tool calls, executing block-dangerous.sh, owner-guard.sh, ritual-log.sh, lessons-entry-budget.py and star-topology.sh with --hook (6 commands). From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs to expand your prompt before the agent sees it, executing ritual-log.sh. From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs when you submit a prompt, before the agent sees it, executing ritual-log.sh. From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs when a session starts on startup, resume, clear and compact, executing session-start-reground.sh, team-addresses.py, lessons-entry-budget.py and ritual-log.sh with --hook and --check (4 commands). From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs before the context is compacted for manual tool calls, executing pre-compact-snapshot.sh, ritual-log.sh and compact-gate.sh (3 commands). From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Hook Claude Code
Runs when the agent finishes a response, executing handover-reminder.sh and plan-phase-nudge.sh (2 commands). From muratsilahtaroglu/claude-code-starter-kit.
muratsilahtaroglu/claude-code-starter-kit
Settings file Claude Code
Agent settings declaring 6 hook events (PreToolUse, UserPromptExpansion, UserPromptSubmit, SessionStart) and 7 allowed tools.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Adopt Keel's discipline into an EXISTING project (brownfield) — inventory, non-destructive merge, back-fill docs from real code, migrate security. Never overwrites.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
OWNER-only setup wizard for a same-machine agent team — name the roster (one orchestrator + specialized workers), generate .claude/agents/team- .md charters with owner-approved full text, seed TASKS lanes + reports/team author folders. Structured co-agents per rules §10.42.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Adopt an agent-team identity for THIS chat (run once per new chat) — load the charter, record the session→agent mapping so the reground hook re-injects the identity from disk after every compaction/--resume, then resume from the agent's lane.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Periodic rules-compliance spot-check — spawns the read-only auditor subagent over the commits since the last audit, reports violations with file:line + rule §, fixes only what the user approves. Skips itself on an empty (phase-0) project.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Gated autonomy for one session — advance phases back-to-back with the full /keel-phase-review run at every gate, auto-commit at phase ends, batch pushes for ONE approval. Stops on gate FAIL, UNCERTAIN decisions, architecture surprises, or security-adjacent changes. Push is never automatic. Solo sessions or an agent…
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Pre-compact ritual — bring the disk fully up to date (runs the /keel-handover procedure, which cascades to /keel-distill, PLAN flips and the LESSONS sweep), verify freshness, then hand off to the built-in /compact. One command to remember instead of several.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Work out what this session should do NEXT and continue from it — resume half-done work, or pick up the lane's newly assigned task, or report that there is nothing to do and go idle. Cross-checks the auto-loaded memory against PLAN.md + git. Role-aware (worker vs orchestrator) and read-only, so any session type may run…
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Consolidate project memory — rotate old HANDOVER blocks to the archive, promote critical facts to LESSONS.md, dedup/merge, and lint for contradictions. Run when caps are exceeded.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Add this session's block to HANDOVER.md before ending or compacting — done / tried-failed / latest / next. Rotates via /keel-distill when the 3-block cap is hit.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
End-of-phase gate — verify working product, tests, docs, ADRs, and handover before committing a phase.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Staged bulk-run gate — before any mass-output job (labeling, generation, migration) enforce, in order, declare thresholds → smoke sample → gold-set validation → progressive ramp with halt conditions → acceptance sample. No full run on an unvalidated pipeline.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Create or revise PLAN.md — propose the phase DAG (phases · gates · dependencies) from the user's goal, get approval, write the table and regenerate the colored Mermaid diagram from it. Statuses flip at rituals (/keel-handover, /keel-phase-review); post-completion fixes land in the Fix log.
muratsilahtaroglu/claude-code-starter-kit
Skill Claude CodeCodex
Opt-in external research — scan GitHub/articles/LinkedIn/HuggingFace/web and save cited findings per platform under research/.