| name | manage-ideas |
| description | Sync LAMBDA ideas between Notion and GitHub Issues, refresh the Current Lambdas catalogue in Notion, and close idea issues whose LAMBDA has shipped. Usage: /manage-ideas |
| allowed-tools | Bash, Read, Glob, Grep, mcp__claude_ai_Notion__notion-fetch, mcp__claude_ai_Notion__notion-search, mcp__claude_ai_Notion__notion-create-pages, mcp__claude_ai_Notion__notion-move-pages, mcp__claude_ai_Notion__notion-update-page |
Manage Ideas — lambda-boss
Keep the Notion ⇄ GitHub idea bridge in sync, in both directions, doing the least work necessary:
- Inbound: sync new LAMBDA ideas from Notion to GitHub Issues.
- Outbound: refresh the "Current Lambdas" Notion catalogue with the LAMBDAs shipped in this repo.
- Reconcile: close
lambda-idea issues whose LAMBDA has just shipped.
The repo is version-controlled, so git is the source of truth for what changed. The catalogue page stamps the commit it was last built from; each run diffs against that commit and only touches what moved. A run where nothing changed and no new idea exists should do almost nothing.
Key Notion page IDs
| Page | ID |
|---|
| Excel eSports (parent) | 3407b3d23d2f80e49224f99b577224db |
| LAMBDA ideas (inbox) | 3407b3d23d2f80aa9ac4e4903868d7c8 |
| Current Lambdas (catalogue) | 3437b3d23d2f819595bac162a5268640 |
| Synced to GitHub (archive, child of LAMBDA ideas) | discovered at runtime — created if missing |
Step 0 — Detect what actually needs doing
Run these cheap probes first, then run only the parts that have work.
-
Fetch the catalogue page once (used for the sync marker and as the row cache for delta rebuilds):
notion-fetch id: 3437b3d23d2f819595bac162a5268640
Extract the last-synced commit SHA from the header line:
_Last synced: <YYYY-MM-DD> · commit `<short-sha>`_
If the header has no commit SHA (e.g. a catalogue built by the old version of this skill), treat the outbound delta as a full rebuild (Step A, full path).
-
Compute the lambda delta against that SHA:
git rev-parse --short HEAD
git diff --name-status <last-synced-sha> HEAD -- lambdas
Keep only *.lambda paths. Classify: A = added, M = modified, D = deleted, R… = rename (treat as delete-old + add-new). If <last-synced-sha> is missing or git diff errors (unknown SHA), fall back to a full rebuild.
-
Probe the inbound inbox for unsynced ideas:
notion-fetch id: 3407b3d23d2f80aa9ac4e4903868d7c8
Count child <page> elements, excluding the "Synced to GitHub" archive container.
Decide the plan:
| Condition | Run |
|---|
| No lambda delta and no SHA-missing fallback | Skip Step A (report "catalogue unchanged") |
| Lambda delta present (or SHA missing) | Step A |
Any added (A) lambdas | Step B (reconcile) |
| Any unsynced child pages | Step C (inbound) |
State the plan in one line before doing the work, e.g. "2 lambdas changed, 1 added, 1 new idea → refresh catalogue, reconcile 1, sync 1."
Step A — Refresh "Current Lambdas" (delta rebuild)
The catalogue is grouped by library (H2), each with a three-column table: Name | Description | Calculation. You will reassemble the whole page and replace_content it, but only re-derive the rows that changed — unchanged rows are reused verbatim from the page fetched in Step 0.
A1 — Reuse unchanged rows
From the catalogue markdown fetched in Step 0, parse the existing rows keyed by LAMBDA name. Every lambda not in the delta keeps its existing row text exactly as-is. Do not re-read its .lambda file.
(Full-rebuild fallback: when there is no SHA marker, skip A1 — every lambda is "changed", so derive all rows in A2 as the original skill did.)
A2 — Derive rows for the delta only
For each added/modified .lambda file in the delta, read it and extract:
-
Name — the function name from the FUNCTION NAME: header.
-
Description — the one-line blurb inside the /**...*/ JavaDoc comment on line 2.
-
Calculation — the meaningful formula logic from the // Procedure section, excluding:
- the
Help TEXTSPLIT block,
- the
// Check inputs bindings (Help?, ISOMITTED(...) guards, default-value normalisers like IF(ISOMITTED(x), default, x)),
- the final
IF(Help?, Help, result) return.
Collapse to a compact inline formula. Intermediate LET bindings can be inlined or described in prose (e.g. "… where _count = CEILING(LEN(text)/_size, 1)"). Aim for a single backticked expression per row; prose only when inlining would hurt readability. Prefer fidelity over prettiness — the reader is checking for duplicate logic, so the formula must be recognisable, not necessarily runnable.
For each deleted/renamed-away lambda, drop its row.
For a library that gains its first lambda, or whose folder is new, read lambdas/<library>/_library.yaml for its name and description. (Library headers/descriptions for unchanged libraries are reused from the fetched page.)
A3 — Reassemble and replace
Build the full page: one H2 per library (sorted alphabetically), library description in italics, rows sorted alphabetically by LAMBDA name. Header block:
Catalogue of LAMBDAs already implemented in [TagloGit/lambda-boss](https://github.com/TagloGit/lambda-boss). Check this list before proposing new LAMBDA ideas to avoid duplicates.
_Last synced: <today YYYY-MM-DD> · commit `<short-sha from git rev-parse>`_
The <short-sha> must be the git rev-parse --short HEAD value from Step 0 — that is what the next run diffs against. Then:
notion-update-page
page_id: 3437b3d23d2f819595bac162a5268640
command: replace_content
new_str: <the reassembled markdown>
Use replace_content (not update_content) — the page is the source of truth and is fully rewritten each run.
Step B — Reconcile shipped ideas (close implemented idea issues)
Run only when the delta contains added (A) lambdas — a newly shipped LAMBDA may be the implementation of an open idea issue that was never linked with Closes #NN.
-
List open idea issues:
gh issue list -R TagloGit/lambda-boss --label lambda-idea --state open --json number,title,body
-
For each newly added LAMBDA, judge whether it implements one of those open issues — match on function name and intent (the issue title is LAMBDA idea: <name>; compare against the LAMBDA's name and description). Be conservative: only act on a confident match.
-
For a confident match, close the issue and mark it done:
gh issue edit <NN> -R TagloGit/lambda-boss --add-label "status: done" --remove-label "status: backlog"
gh issue close <NN> -R TagloGit/lambda-boss --comment "Shipped as LAMBDA \`<NAME>\` (\`lambdas/<lib>/<NAME>.lambda\`). Closing the idea."
(Remove whichever status label the issue currently carries — usually status: backlog.)
-
For an uncertain match, do not touch the issue; list it under "Possible matches — needs Tim" in the report.
Step C — Sync ideas inbound (Notion → GitHub)
Run only when Step 0 found unsynced child pages.
C1 — Ensure the "Synced to GitHub" archive exists
From the inbox fetch in Step 0, check for a child page titled "Synced to GitHub". If absent, create it and save its ID for C4:
notion-create-pages
parent: { type: "page_id", page_id: "3407b3d23d2f80aa9ac4e4903868d7c8" }
pages: [{ properties: { "title": "Synced to GitHub" }, content: "Archive of LAMBDA ideas that have been synced to GitHub Issues." }]
C2 — Cheap duplicate check (names, not the catalogue)
Build the set of existing LAMBDA names from a filename Glob — lambdas/**/*.lambda, take each basename without extension. This is all the dedup check needs; do not rebuild or re-read the catalogue for this. (Step A, if it ran, already gives you the names too.)
C3 — For each unsynced idea page
notion-fetch id: <page-id>
Duplicate check — compare the idea's title/intent against the existing LAMBDA names from C2. If it clearly duplicates an existing LAMBDA:
- Leave the Notion page where it is (do not archive it).
- Add it to the "Skipped (duplicates)" section of the report, noting the LAMBDA it matches.
Otherwise, create the GitHub Issue:
gh issue create -R TagloGit/lambda-boss \
--title "LAMBDA idea: Weighted Score Calculator" \
--label "enhancement" --label "lambda-idea" --label "status: backlog" \
--body "$(cat <<'EOF'
## Origin
Synced from Notion: [LAMBDA: Weighted Score Calculator](https://www.notion.so/...)
## Content
...
EOF
)"
C4 — Move synced page to archive
After the issue is created successfully:
notion-move-pages
page_or_database_ids: ["<page-id>"]
new_parent: { type: "page_id", page_id: "<synced-to-github-page-id>" }
Step D — Report results
Summarise only what ran:
Catalogue: refreshed (3 rows changed, 1 added, 0 removed) at commit a1b2c3d.
— or — Catalogue: unchanged (no lambda changes since a1b2c3d).
Reconciled (closed implemented idea issues):
- TagloGit/lambda-boss#NN "Weighted Score Calculator" — shipped as WSCORE
Synced N idea(s) from Notion to GitHub:
- "Weighted Score Calculator" -> TagloGit/lambda-boss#NN
Skipped as duplicates:
- "Running Total" — matches existing CUMSUMGRID
Possible matches — needs Tim:
- TagloGit/lambda-boss#MM "Foo" — maybe shipped as FOOBAR? (unsure)
If a section is empty, omit it. If nothing ran at all: "Nothing to do — catalogue current, no new ideas."
Idempotency
- The catalogue is rebuilt from git state and stamped with HEAD's SHA; a re-run with no new commits finds an empty delta and skips Step A.
- Ideas already moved to "Synced to GitHub" don't appear as inbox children next run.
- The "Synced to GitHub" page is created only if missing.
- Reconciliation only closes issues on a confident match and is a no-op once they're closed.
Behaviour Notes
- Do NOT modify the content of idea pages — sync them as-is.
- Do NOT create specs or plans from ideas — they go to the backlog for triage.
- Strip the "LAMBDA: " prefix from idea titles when creating issues (use "LAMBDA idea: ").
- If a Notion page fetch fails, skip it and report the error — don't stop the whole sync.
- If parsing the existing catalogue for row-reuse fails for a given library/row, re-derive that row from its
.lambda file rather than dropping it.
- The first run after this skill update will find no SHA marker → one full rebuild that plants the marker; every run after is incremental.
