From beislid
Starts ticket work after feature branch checkout: fetches ticket, explores codebase, routes vague/large tickets to spec/break-spec, then blueprint/implement. Driven by .beislid/workflow.md.
npx claudepluginhub sandsower/beislidThis skill uses the workspace's default tool permissions.
Turn a checked-out feature branch and ticket into an implementation-ready plan. Kickoff is the front door of the Beislið lifecycle: ticket → context → spec/blueprint/breakdown → ticket update → `implement`.
Interactively configures Beislið project workflow in .beislid/workflow.md using git prechecks, ticket source detection, and conversational menus. Shows diffs before writes; activates on setup/configure intents.
Executes plan files or Beads epics step-by-step: git branch/worktree setup, phase-by-phase task tracking, quality checks, tests, commits, optional code review.
Drives end-to-end ticket workflow: analyzes tickets, investigates codebase, plans, implements via TDD, commits changes, and creates PRs. Adapts to stack via interactive supplement setup.
Share bugs, ideas, or general feedback.
Turn a checked-out feature branch and ticket into an implementation-ready plan. Kickoff is the front door of the Beislið lifecycle: ticket → context → spec/blueprint/breakdown → ticket update → implement.
Don't use this for: creating PRs, handling review/QA feedback, or shipping completed work. Use ship-it or heard-chef for those.
Project config lives at <repo>/.beislid/workflow.md (typed-key fenced YAML blocks; format reference at workflow-md-format.md). Capabilities are probed lazily on first need per probe-semantics.md. Output prose follows output-templates.md and kickoff-templates.md.
Read <repo>/.beislid/workflow.md. If missing, hard-fail and stop:
🛑 No
workflow.mdfound in.beislid/. If this is a fresh project, run/setupto create one. If you moved your config, restore it to<repo>/.beislid/workflow.md.
Validate line 1 is exactly <!-- beislid-workflow: v1 -->. If missing or wrong, hard-fail and stop; kickoff will not silently mis-parse.
Compute cache identifiers:
repo_hash=$(git rev-list --max-parents=0 HEAD | sort | head -c 12)
workflow_hash=$(git hash-object .beislid/workflow.md)
Try to read ${BEISLID_STATE_DIR:-$HOME/.local/state/beislid}/probes/<repo_hash>.json:
cold.workflow_hash mismatch → empty in-memory probe state; cache state stale.fresh. Per-cap freshness uses probe_cache.ttl_hours, default 24.Print orientation prose from kickoff-templates.md (≤240 chars).
Reusable lazy-probe contract. Do not re-probe within a run; the in-memory probe state is authoritative.
<cap> in memory.status: ok, and within TTL → return ok; no probe ran.probe-semantics.md for the cap kind (mcp, cli, file, paste, skill, subagent, path). Record the result in memory.kickoff-templates.md with retry / proceed-this-session fallback / abort.session_skip: true; exclude from run-end writeback. On abort → stop and do not write cache.On clean run end, write probed/re-probed capability entries back to <repo_hash>.json, excluding session_skip. Preserve top-level fields doctor owns; orchestrators do not update doctor_run_at.
.beislid/workflow.md is the only project config source; no legacy YAML fallback.{body_file}, never raw body interpolation.Complete steps in order. At each step entry, read the step aux file and follow it as the authoritative protocol. Do not execute a step from memory if the aux read fails; hard-fail and stop:
🛑 Could not read
skills/kickoff/<step-file>.md. Kickoff cannot safely execute this step from memory; reinstall Beislið or restore the file.
When BEISLID_VERBOSE=1, print one aux load stamp after successfully reading a step file, e.g. ✓ kickoff/step-1-ticket v1 loaded, and include step exit checks in the transcript/output when practical.
step-1-ticket.md. Outputs: ticket_id, title, body, acceptance criteria, labels/metadata, attachments/screenshots summary, ticket-source status.step-2-context.md. Outputs: relevant files/patterns/tests/docs, domain context status, open uncertainties.step-3-team-guidance.md. Outputs: team config status and constraints.step-4-readiness.md. Outputs: route to spec or blueprint path, with rationale.step-5-scope.md. Outputs: single PR or needs breakdown, with selected phase if any.step-6-blueprint.md. Outputs: approved design summary, expected files/modules, tests, risks/open questions.step-7-discoveries.md. Outputs: discovery status and any durable notes recorded.step-8-ticket-update.md. Outputs: update status/body and implement handoff context.After Step 8 or a clean routed handoff, write in-memory probe state to <repo_hash>.json: update probed/re-probed entries, exclude session_skip: true, preserve doctor_run_at, update workflow_hash, and keep the workflow TTL. If workflow.md changed mid-run or the run aborted, do not overwrite stale state.
knowledge_store.path as useful alone.