Combine the interview-driven grill and the domain-modeling discipline into a
single model-invokable grill skill, so other skills can reach it as a backbone.
- One question at a time, each with options plus a recommendation and rationale,
and a context quote when it turns on existing material.
- Agent-agnostic mode selection: discover .agents/CONTEXT.md then
.claude/CONTEXT.md, and drive Domain Modeling, Free, or Clarification.
- Domain Modeling Mode maintains the glossary inline (CONTEXT-FORMAT.md) and
offers ADRs under the three-gate test.
- Never begin implementation without an explicit user instruction, even when a
caller skill invoked grill.
Both spec docs and tasks 0001-0008 were mined into the personal wiki by
/consume; delete the scaffolding it consumed, leaving the now-empty spec/
and tasks/ directories in place.
Reworks the consume skill around two acceptance axes and two destinations,
implementing the six issues filed from the first real consume run as one design.
- Add trigger-ability as a second acceptance axis alongside generalizes, with
its two premises; a pull-channel note now requires both (#9).
- Give consume a push channel — the target's own CLAUDE.md — for knowledge that
fails an axis but is still useful: specific residue or proactive rules (#10).
- Name the general-lesson/specific-residue split and route each half (#12).
- Make the step-5 plan report a complete ledger with a per-note trigger line,
push write-set, ref-fixes, flagged-not-authored, and drops (#13).
- Extend project cleanup to reconcile CLAUDE.md (capture-gated removal/trim) and
fix references to deleted files wherever they land (#11).
- Encode a branch-owned output location: project notes flat under
01 sources/claude/projects/, honoring any pre-existing grouping (#14).
Records the resolved vocabulary in .claude/CONTEXT.md and the two-channel write
decision in .claude/adr/0001.
Closes#9Closes#10Closes#11Closes#12Closes#13Closes#14
/consume is the write-side counterpart to the read-only /wiki: it mines a
target for generalized, transferable knowledge and files atomic concept
notes into 01 sources/claude/, the agent-owned area of the Obsidian vault.
It is user-invoked/deliberate. SKILL.md holds the branch-detection ladder
plus the common writing logic — read existing knowledge (via /wiki and a
direct scan), plan atomic generalized notes with a file-to-section-to-note
contribution map, present an HTML plan and wait for explicit approval, then
write on the vault's note template and clear the consumed scaffolding.
v1 ships one branch, project.md: source code is truth for current state,
while .claude/spec and .claude/tasks are the record of why (mined, then
deleted). The ladder is extensible for future source types.
A model-invoked skill that answers questions from the user's Obsidian
vault, read-only, over the filesystem. Enforces the vault's own
settings.json deny-Read globs itself, since the harness does not apply
them when running from another directory. Uses the 02 tags/ hubs as a
recall backstop and reads to saturation, then answers with per-note
provenance.
Add a second pure transform to the benchmark core: a `--leaderboard` mode
that ranks per-skill results models into an index leaderboard, one row per
skill carrying both verdicts, links out to each per-skill report, and sorts
any red first (regressions above efficacy failures), then fragile-but-passing,
then clean green.
The fragile tier reuses the 0006 per-badge chips: the efficacy chip now
carries its win count so the leaderboard reads it against the pass floor
without recomputing margins, and fragility is scoped to the passing tier so a
red row never carries a chip. The runner's SKILL.md gains the no-argument
batch flow and the index invocation. The fixture test covers the tiered sort,
the not-applicable Regression cell, and the per-skill links.
Add the longitudinal layer to the benchmark core: a per-skill JSON-lines
history that appends one summary line per run and trims oldest-first at
fifty, two stacked net-margin trend ribbons (Efficacy and Regression, the
latter leaving a gap for two-arm runs) with the current run ringed and a
net/delta/green-count readout, and per-badge fragility chips that flag the
narrowest passing case on each green axis so a barely-green skill cannot
look robust.
History persistence is opt-in via --history and is the core's only side
effect. Without it the core stays a pure transform. Two committed fixtures
(clean and fragile) and new fixture-test sections cover the append-and-trim,
both ribbons including the two-arm gap, and the chips.
