<< All versions
Skill v1.0.2
currentAutomated scan100/1002lab-ai/oh-my-claude/what-we-have-to-work
~1 modified
──Details
PublishedAugust 6, 2026 at 01:48 PM
Content Hashsha256:41edce87326b06c5...
Git SHA844816914697
Bump Typepatch
──Files
Files (1 file, 3.5 KB)
SKILL.md3.5 KBactive
SKILL.md · 74 lines · 3.5 KB
version: "1.0.2" name: what-we-have-to-work description: "Use when unfinished trace scenarios exist and the next execution scope must be chosen. Bundles leftovers of ANY size into 1-3 options and hands the selected bundle contract (trace_path + scenario_ids) to do-work."
What We Have To Work
Goal
Turn unfinished scenarios from docs/*/trace.md into up to three clear work bundles and get a fast user selection.
Decision Gate (MANDATORY)
Read `${CLAUDE_PLUGIN_ROOT}/prompts/decision-gate.md` and apply it to every decision in this workflow.
Sizing Rubric: read ${CLAUDE_PLUGIN_ROOT}/prompts/decision-gate.md (single source — do not duplicate the table here). Sizes below (tiny/small/medium/large/xlarge) refer to expected code change (added + deleted).
Workflow
- Survey trace scenarios
- Glob for
docs/*/trace.mdin the project - Read Implementation Status table from each trace.md
- Collect unfinished scenarios with their feature, dependencies, and size estimate
- Note which features have specs (docs/{feature}/spec.md)
- Shape the backlog into bundles
- Every unfinished scenario is execution-eligible — small and medium tails included. Do NOT reroute to plan-new-task just because the backlog is small.
- Target large/xlarge bundles when the backlog allows; a single small/medium leftover is still a valid bundle on its own.
- Never exceed xlarge per bundle.
- Build bundles
- Create one to three bundles, not more
- Do not force three bundles if the backlog is small
- Aim for xlarge bundles; large is acceptable if xlarge would require unrelated work
- Do not exceed xlarge
- One bundle = one trace.md. The bundle contract carries a single
trace_path, so a bundle can never span features. Within that one trace, group scenarios by dependency chain or shared code area. - Genuinely coupled work across two features → two bundles, presented with an explicit execution order (never one merged bundle)
- Attach tiny leftovers as add-ons to related bundles when possible; if nothing related exists, a leftover-sweep bundle of small scenarios is valid
- Do not mix unrelated features just to hit size targets
- Present options
text
I found {N} unfinished scenarios across {M} features. Here are the bundles:1) {Bundle title} ({size tier})Feature: {feature-name} (docs/{feature}/trace.md)- Scenario {n}: {title}- Scenario {m}: {title}Rationale: {short reason}Contract: {"trace_path": "docs/{feature}/trace.md", "scenario_ids": ["n", "m"], "size": "{tier}", "rationale": "{short reason}"}2) ...3) ...Reply with 1, 2, or 3 and any extra context.If you want a different bundle, say what to include.
- After user selection
- Confirm the selected bundle
- Invoke
Skill(skill="stv:do-work")passing the selected bundle contract verbatim —trace_pathandscenario_idsMUST survive the handoff unchanged; do-work must not rescan the project when this contract is provided. - Validate before handoff:
trace_pathresolves to an existing trace file and everyscenario_idsentry exists in that trace's Implementation Status; on mismatch, stop with a targeted-scope error instead of handing off.
- Empty backlog
- Only if ZERO unfinished scenarios exist, stop and invoke
Skill(skill="stv:plan-new-task").
Integration
- Use
stv:what-to-workfor routing decisions (it calls this skill) - Use
stv:plan-new-taskonly when zero unfinished scenarios remain - Use
stv:do-workafter the user picks a bundle