Study status
Opinionated Agent Skills for AI-native autodidacts — deep learning, dense papers, public proof. Opinionated Agent Skills for AI-native autodidacts — deep learning, dense papers, public proof. By School of the Bear
npx -y skills add loursCreatif/ours-stack --skill study-statusAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Read-only study dashboard — where each study stands (criteria checked, artifacts produced, suggested next step). Scans studies/*/brief.md in the current repo; optionally includes ~/.ours-stack/studies-index.jsonl for cross-repo studies. Use when the user asks "study status", "où j'en suis", "statut de mes études", "tableau de bord", "what's next", "what should I do next" (status view, not strategy debate), or runs /study-status. Never creates or modifies files.
SKILL.md
8.3 KB, as published. Nobody here has run it
Study Status
Read-only dashboard — one glance at every local study: wedge, success criteria progress, artifacts on disk, and the next skill the routing table suggests. No file writes, no research, no summary theater.
Time budget: ~1–2 minutes. Scan briefs + artifact presence only — do not read notes.md or anchor sources in depth.
Hard rules
- Read-only — never
Write,Edit, or create files. Output lives in chat only. - No summary theater — table + 2–3 bullets. No essay, no "great progress!" filler.
- Routing fidelity — next-step suggestions must follow
AGENTS.mdrouting (see below). Do not auto-chain skills. - Wedge in one line — truncate
## Narrow wedgeto ≤80 chars for the table; full wedge only if user asks. - AskUserQuestion — one question per call; title
Study Status — {label}.
When to use (vs other skills)
| Skill | Role |
|---|---|
/study-status | This skill — status table + next step per study |
/council | Live round table — debate the topic with a multi-figure panel |
/source-scout | Find sources — does not report status |
/bear-hours | Frame a new study — not a dashboard |
Not auto-chained from any other skill.
Step 0: Scope
Default: current workspace — Glob studies/*/brief.md.
If no briefs found → say so; suggest /bear-hours for a new topic. Stop unless user names another repo path.
Optional global index — after listing local studies, AskUserQuestion:
Title: Study Status — Cross-repo?
Ask: "Inclure les études d'autres repos via ~/.ours-stack/studies-index.jsonl ?"
Options:
- Oui — afficher slug, titre, repo, chemin (sans ouvrir les briefs externes)
- Non — repo courant seulement
other
If yes → Read ~/.ours-stack/studies-index.jsonl (or $OURS_STACK_HOME/studies-index.jsonl if set). For each line, parse JSON fields: slug, title, wedge, repo, path. Skip entries whose path is already in the local scan. Do not open external briefs except top 2–3 most relevant (same wedge keyword overlap or user-named slug).
If user names one slug → scan only studies/<slug>/brief.md (still list others only if they ask "all studies").
Step 1: Discover studies
# List local study folders with briefs
Glob studies/*/brief.md
For each match, record slug from the path (studies/<slug>/brief.md).
Step 2: Extract brief fields
Read each studies/<slug>/brief.md. Extract:
| Field | How |
|---|---|
| Title | First # heading line |
| Slug | **Slug:** line, or dirname |
| Created | **Created:** line (ISO date) |
| Time box | **Time box:** line (free text) |
| Wedge | First substantive line under ## Narrow wedge (bold unit line preferred) |
| Criteria | All - [ ] and - [x] under ## Success criteria — count checked/total |
| Sources | ## Source material — TBD if section missing, empty, or contains only "TBD" / "à scout" / placeholder; OK if anchor + core list present |
Criteria parsing: only lines matching ^- \[[ x]\] inside ## Success criteria until the next ## heading.
Source material heuristic:
| State | Condition |
|---|---|
| TBD | Section absent, or only placeholders ("TBD", "à définir", "run /source-scout") |
| OK | Has ### Read first or ### Core or ### Anchor with at least one linked source |
| Partial | Some sources listed but no clear anchor — note in table, suggest /source-scout refresh |
Step 3: Detect artifacts
For each studies/<slug>/, Glob or Bash test -f / test -d against the AGENTS.md contract:
| Artifact | Path | Label |
|---|---|---|
| Notes | notes.md | notes |
| Council | council.md | council |
| Dialogue | dialogue/*/ (any subfolder with dialogue.md) | dialogue |
| Layout | report.html | html |
| Mind map | mind-map.json or mind-map.html | mind-map |
| Infographic | visual-proof.png, visual-proof.spec.md, or infographic-prompt.md | visual |
| Memory palace | memory-palace.json or memory-palace.html | palace |
| Deep research link | research.md or sources-index.md | research |
Display: compact comma-separated labels in the Artefacts column (e.g. notes, council, mind-map). Use — if none beyond brief.md.
Do not count brief.md as an artifact.
Step 4: Time box signal
Parse Created as YYYY-MM-DD. Interpret Time box heuristically:
| Time box text | Deadline estimate |
|---|---|
| "aujourd'hui", "one session", "today" | Created + 0 days |
| "une semaine", "1 week" | Created + 7 days |
| "deux semaines", "2 weeks" | Created + 14 days |
| explicit date in text | use that date |
| unclear | skip overdue check |
If estimated deadline < today (from user_info Today's date) → append ⏰ to the study row in Prochaine étape and mention in bullets ("time box dépassée").
Step 5: Suggest next step (routing)
Apply in order per study — first match wins:
| Condition | Suggested step |
|---|---|
No brief.md | /bear-hours |
| Sources TBD or Partial | /source-scout |
Sources OK, no notes.md | /dense-read (anchor from brief) |
notes.md present, criteria not all checked | Continue /dense-read — or /council to debate the topic if stuck |
| All criteria checked | Étude terminée — archiver, ou nouveau wedge via /bear-hours |
| User asked "what should I do next?" on one study with real ambiguity | This skill shows facts first; suggest /council only if the user wants to debate the topic |
Optional artifacts (never auto-suggest unless user history implies it): /mind-map, /layout-html, /infographic, /memory-palace, /dialogue — mention in bullets only if criteria nearly done and no visual artifact yet.
Step 6: Output format
Main table
| Étude | Wedge | Critères | Artefacts | Prochaine étape |
|-------|-------|----------|-----------|-----------------|
| robotique-assemblage-structurel | comment un robot pose des éléments structurels… | 0/3 | — | `/dense-read` |
| biomimetisme-locomotion-chantier | locomotion pattes sur terrain instable… | 2/3 | notes, council, mind-map, visual, palace | `/dense-read` |
- Étude column: slug (link path
studies/<slug>/brief.mdif the UI supports it, else plain slug). - Keep wedge ≤80 chars with
…if truncated. - Prochaine étape: one skill or short phrase; add
⏰prefix if time box overdue.
Cross-repo appendix (if index included)
Separate small table — no wedge/criteria unless brief was opened:
| Slug | Titre | Repo | Chemin |
|------|-------|------|--------|
Synthesis bullets (2–3 only)
Pick concrete facts, e.g.:
- Closest to done:
<slug>— N/M criteria, artifacts: … - Stalled:
<slug>— sources TBD since Created, or time box ⏰ - Dormant: studies with Created >14 days ago and no
notes.md
No motivational language. No recommendations beyond the routing table.
Example invocation
/study-status
/study-status robotique-assemblage-structurel
/où j'en suis sur mes études
Failure modes
| Situation | Response |
|---|---|
No studies/ in workspace | "Aucune étude locale. Lance /bear-hours pour cadrer un sujet." |
| Slug not found | List available slugs from Glob; ask once |
studies-index.jsonl missing | Note file absent; local scan only |
| Malformed brief (no wedge) | Show row with wedge —; suggest brief cleanup via /bear-hours revisit |
Privacy
Status output may echo wedge text and criteria from briefs. Do not paste long quotes from notes.md or copyrighted source titles beyond what the brief already lists.