| name | second-brain-ideate |
| description | Mine the knowledge-base wiki for strong, defensible content ideas — blog posts, conference talks, internal sessions, threads. Use when the user says "ideate", "what should I write about", "find content ideas in my wiki", "blog/talk ideas", or wants to turn collected knowledge into publishable content. Produces a scored shortlist with outlines in output/ and maintains an ideas backlog in wiki/synthesis/.
|
| allowed-tools | Bash Read Write Edit Glob Grep |
Second Brain — Ideate
Extract content ideas from the wiki's structure between pages — collisions, contradictions, authority clusters, gaps — not from single pages.
Grounded ideation: every idea must cite the wiki pages that back it, so the output is a plan, not a brainstorm.
Phase 1 — Harvest signals (mechanical)
Collect the raw material cheaply before any judgment:
grep -roh '\[\[[^]|]*' wiki/ | sed 's/\[\[//' | sort | uniq -c | sort -rn | head -30
grep -l "sources: \[.*,.*\]" wiki/concepts/*.md wiki/entities/*.md wiki/synthesis/*.md 2>/dev/null
grep -ril "contradict\|disagree\|conflicting" wiki/concepts wiki/synthesis wiki/entities 2>/dev/null
grep -h "^tags:" wiki/*/*.md | sort | uniq -c | sort -rn | head -20
tail -60 wiki/log.md
Also read wiki/index.md (and wiki/synthesis/content-ideas-backlog.md if it exists — never re-propose an idea already marked picked/drafted/published/dropped).
Read the user's own authored sources (deep pages sourced from their blog) — personal authority is a scoring input.
Phase 2 — Generate candidates (five patterns)
Walk each pattern against the harvested signals; a candidate is only valid with 2+ supporting wiki pages attached:
- Collision — two clusters sharing a structural pattern that never cite each other (e.g. an infra maturity model × AI-agent autonomy). Highest yield: the credibility is knowing both sides.
- Contradiction — sources in the vault that disagree; the piece names the variable that decides who's right.
- Authority intersection — concepts with multiple sources plus the user's first-person experience (their own deepened posts, their log activity). These survive Q&A.
- Gap — a question the wiki raises that no source answers, especially where the user's recent work is an answer.
- Fresh × evergreen — a recent clip that lands on a durable concept already in the vault.
Read the actual candidate pages (not just titles) before keeping a candidate — the connection must be real, not name-similarity.
Phase 3 — Score
Score each candidate 1–5 on four axes; drop anything under 12 total:
- Novelty — would this combination be hard for someone else to write? (Collisions and gaps score high; summaries of one source score 1.)
- Evidence — how many wiki pages back each beat of the argument?
- Authority — does the user have first-person experience or authored sources here?
- Audience fit — is there an obvious venue (their blog's beat, a conference CFP, an internal session, a thread)?
Recommend a format per idea: blog post, conference talk, internal session, or thread — match depth of evidence to depth of format (talks need authority ≥4).
Phase 4 — Output
- Write the report to
output/content-ideas-YYYY-MM-DD.md: top 5–10 ideas, each with
- working title, format, one-paragraph pitch
- the 3-beat argument, each beat citing its supporting
[[wiki pages]]
- score breakdown and the pattern that produced it
- Create or update
wiki/synthesis/content-ideas-backlog.md: one line per idea with status new | picked | drafted | published | dropped; carry forward prior entries untouched.
Frontmatter per wiki conventions (tags, sources it draws on, created/updated).
- Append to
wiki/log.md:
## [YYYY-MM-DD] ideate | N ideas (M new, K carried)
Top: "Working Title A" (collision, 18), "Working Title B" (gap, 16). Backlog: [[Content Ideas Backlog]].
- Report the shortlist to the user and ask which idea to pick; on pick, mark it
picked in the backlog.
Conventions
- Ideas are grounded or they don't ship: no candidate without named wiki pages behind every beat.
- The backlog page is the memory across runs; the report in
output/ is the disposable artifact.
- Composes with the rest of the loop:
/second-brain-ingest deepen the supporting pages of a picked idea first, then draft (e.g. with a blog-writing skill) from the beats.
- Respect vault roles: read
wiki/, write reports to output/, the backlog is the only wiki page this skill maintains.
Related Skills
/second-brain-query — explore a topic's coverage before or during ideation./second-brain-ingest — deepen the supporting pages of a picked idea before drafting./second-brain-lint — fix broken links/orphans first; a healthy graph yields honest signals.
