Skip to content
FSS Knowledge Base

Ideas-Workflow — Roadmap

Section titled “Ideas-Workflow — Roadmap”

Multi-phase plan from zero-friction capture to curated briefings. Phases run in order; each phase gets /task-prep at phase start (not in advance — prepped context goes stale) with a model recommendation baked in.

Now

Section titled “Now”

Active task: Ideas-Workflow-P4-multi-format-intake (P4). /task-continue D:\FSS\KB\Core\Processes\Projects\Ideas-Workflow\Tasks\Ideas-Workflow-P4-multi-format-intake.md

P1 — Capture foundation (Web Clipper) ✅ complete 2026-07-17

Section titled “P1 — Capture foundation (Web Clipper) ✅ complete 2026-07-17”

Fix the leak, standardize the clip.

  • Design decisions confirmed by Talbot (2026-07-09/17): single intake Core/Misc/Clippings/; use existing tags prop; add dept prop (Strategy / Mktg / AI / IT, default AI)
  • Web Clipper setup doc + clip template written → Ideas-Workflow-Web-Clipper-Setup
  • Talbot: apply extension settings (browser-side — fixes clips landing in D:\FSS\KB\Clippings\)
  • Verify: test clip lands in Core/Misc/Clippings/ with correct frontmatter
    • tested, working
    • Claude-verified 2026-07-17: Coconote test clip — source, created, tags: clippings, dept: AI all present; no vault-root Clippings/ folder
  • Model: Haiku/Sonnet — config verification only. Task: Ideas-Workflow-web-clipper-capture (closed 2026-07-17)

P2 — Triage v0 (/triage-inbox lite) ✅ complete 2026-07-17

Section titled “P2 — Triage v0 (/triage-inbox lite) ✅ complete 2026-07-17”

The routing mechanism. Shared deliverable with KB-OS M4; skill built in-thread (Talbot’s call).

  • Scope decided: v0-lite (Clippings only) — dept Inboxes deferred to v1; skill arg-shaped for full M4 scope (/triage-inbox [clippings|<dept>|all])
  • Skill built + deployed all surfaces: rate 0–10, TLDR, route KEEP/INBOX/TASK/DISCARD, human-gated batch approve, frontmatter write-back (ai-config c4ebcd9)
  • Acceptance test green: Coconote clip end-to-end (routed INBOX → MBR/Mktg, rated 6, scan 93→92)
  • Key-phrase bolding deferred (UPGRADES) — with dedupe-by-source v0.1 candidate
  • Log: 2026-07-17_Triage-v0-Skill · spec: Ideas-Workflow-Triage-Spec · task: Tasks/archive/2026-07/Ideas-Workflow-triage-v0.md

P3 — Backlog triage (93 existing clippings) ✅ complete 2026-07-17

Section titled “P3 — Backlog triage (93 existing clippings) ✅ complete 2026-07-17”
  • Batch session(s) through the P2 mechanism — 8 batches of ~12, not one-by-one
  • Route/archive per save criteria (Capture-Routing) — 44 kept/routed, 49 archived, 0 remaining
  • Model: Sonnet — mechanical batch work, ran the full backlog in one session.
  • Log: 2026-07-17_P3-Backlog-Triage

P4 — Multi-format intake ⬜

Section titled “P4 — Multi-format intake ⬜”

Borrow > build: research existing tools first; deterministic (non-AI) processing wherever possible.

  • PDFs: deterministic extraction (Stirling-PDF / CLI) — includes Globe *.pdf backlog (see Later/Automations.md)
  • YouTube: _Process playlist → transcript → summary flow
  • Feeds: Medium Daily Digest (email or login scrape), other subscriptions
  • Audio: capture path TBD
  • Web Clipper per-source templates (YouTube Summary, Medium Summary — community templates may exist; requested in Dynalist import)
  • Model: /research skill + Fable for tool-selection plan; Sonnet for build.

P5 — Briefings (“Ideas Dashboard”) ⬜

Section titled “P5 — Briefings (“Ideas Dashboard”) ⬜”

The payoff: curated daily/weekly briefing of what most impacts defined business objectives.

  • Design briefing format (markdown + links on/near Core/DASHBOARD.md — not a “newspaper” HTML page)
  • Web-URL access requirement (Talbot, 2026-07-17): Ideas Dashboard readable anywhere/anytime via URL — like reading Medium Daily Digest AI summaries from Gmail. Options at design time: publish briefing via existing kb-business → Cloudflare Pages pipeline (kb-mbr precedent), emailed digest, or Obsidian Sync (real-time mobile vault access). Idea consumption is non-urgent; briefings may later carry urgent/important items — access design must serve both. Near-term: keep upgrading Obsidian DASHBOARDs (Obsidian Sync would make them mobile real-time).
  • Rank new ideas against Biggest Rocks table; TLDR most important; elevate biggest-rock ideas
  • Also feed from session digests + dept Inboxes (internal intelligence), not just clippings
  • Later: competition monitoring, market/KPI data, dept roll-up briefings
  • Model: Fable for design; Sonnet for the recurring generation job.

Cadence rules

Section titled “Cadence rules”
  • /task-prep each phase at its start, embedding context + model choice (Talbot’s directive, 2026-07-17)
  • Phase-boundary close per SMTM (log to logs/, update STATUS)