Skip to content
FSS Knowledge Base

AI Inbox

Section titled “AI Inbox”

Append-only capture. One idea per bullet. Promote to Tasks/ when ready to work.

Capture

Section titled “Capture”

  • [PHASE-1 FIRST DELIVERABLE] AI infrastructure overview — Document current AI operating setup: Claude Code as primary, model tiers (Opus for judgment, Sonnet for execution), skills library location, graphify setup, cron job monitoring.

  • Skill library audit — run /skills-dashboard to inventory all skills; identify gaps, redundancies, untested skills.

  • Graphify integration — after vault consolidation: run graphify on full D:\FSS\KB, install —watch daemon, add “query graph before grep” convention to AI JOB_DESCRIPTION.

  • .agents/skills/ misplacement flag (from MBR consolidation 2026-07-09): a caveman skill suite lives inside the vault at MBR/_WorkingOn/.agents/skills/ (cavecrew, caveman-commit/-compress/-help/-review/-stats) — decide whether it belongs in ~/ai-config (skills SSOT) or is intentionally vault-local; confirmed for AI-dept review by Talbot.

  • Graphify workflow suggestions (routed from archived Graphify-usage task, full text in MBR/_WorkingOn/Tasks/archive/2026-07/): (1) add /graphify --update to session-close workflow; (2) task files should describe current truth, not stale-reference history; (3) /task-prep enhancement — auto-generate Background via graphify BFS query (~10 min → ~2 min); (4) index the FSS Business KB for cross-vault queries; (5) evaluate --wiki output as an agent entry point (~1k-token orientation); (6) graphify hook install git post-commit auto-rebuild for KB + monorepo; dev-repo inclusion via symlinks (Option A, task-scoped) or merge-graphs (Option B).

  • Skill/plugin research (routed from archived Dynalist export): honest-thinking-partner skill (https://github.com/aiskilloftheweek/claude-ai-skill-of-the-week/blob/main/skills/010-honest-thinking-partner/SKILL.md); evaluate CC plugins for mBR progress — claude-code-setup, Playground, Codebase-scanner; consider a deep-research skill that consolidates findings from multiple AI reports; brainstorm-skill for novel mBR services; hook-writing skill (Mktg).

  • Free official Claude Code course (DeepLearning.AI × Anthropic, Elie Schoppik): clip (triage 2026-07-17, rated 5)

  • Context Engineering vs. Multi-Agent Strategy framework — solo vs. team agent design tradeoffs: clip (triage 2026-07-17, rated 5)

  • AI dev-team case study — cost/architecture breakdown replacing $180K/yr routine dev work: clip (triage 2026-07-17, rated 5)

  • [PROMOTE] 7 ChatGPT writing prompts — Talbot’s own note: automate as Claude Skills substeps: clip (triage 2026-07-17, rated 6)

  • Boris Cherny’s (CC creator) 22 setup/workflow tips — worth auditing our own config against: clip (triage 2026-07-17, rated 5)

  • $10/hr AI-agent delegation math (Sonnet burn rate vs. human contractor rates) — agent-economics reference: clip (triage 2026-07-17, rated 5)

  • Claude Code hooks gotchas from 6mo production — case-sensitive matchers, 50% default activation rate: clip (triage 2026-07-17, rated 6)

  • CC 2.1 features — skill hot-reload, lifecycle hooks, forked sub-agents: clip (triage 2026-07-17, rated 6)

  • Terminal-based AI writing workflow (Claude Code + Warp) — setup worth modeling: clip (triage 2026-07-17, rated 5)

  • Paperclip AI — open-source multi-agent-to-company coordination platform: clip (triage 2026-07-17, rated 6)

  • Anthropic eng lead argues HTML beats Markdown for AI output — worth evaluating for our own report/skill formats: clip (triage 2026-07-17, rated 5)

  • clip rated 7: Which of CLAUDE.md/AGENTS.md/SKILL.md owns what in 2026 conventions; triage 2026-07-18

  • clip rated 7: Full workflow for using Claude to organize/distill an Obsidian vault; triage 2026-07-18

  • clip rated 7: Superpowers 6.0 — self-optimizing, 60% token cut; triage 2026-07-18

  • clip rated 6: Free Gemini CLI setup tutorial, no credit card needed; triage 2026-07-18

  • clip rated 6: Argues most system-prompt content is now redundant with better model defaults; triage 2026-07-18

  • clip rated 6: Token-efficiency technique for Claude usage; triage 2026-07-18

  • clip rated 6: Guardrails for using AI in a PKM without it taking over; triage 2026-07-18

  • clip rated 5: Skill-discovery resource, 50 skills filtered to content-creator use cases; triage 2026-07-18

  • clip rated 5: 8 lesser-known Claude Code slash commands; triage 2026-07-18

  • clip rated 5: One repeated question that speeds up agent task-solving; triage 2026-07-18

  • clip rated 5: Senior PM’s practical Claude Code habits after a year of daily use; triage 2026-07-18

  • clip rated 5: Three specific Claude skills a solopreneur relies on weekly; triage 2026-07-18

  • clip rated 5: Stanford’s STORM method for structured cited research; triage 2026-07-18

  • clip rated 5: Personal task system built on SQLite + Claude Code + GTD; triage 2026-07-18

  • clip rated 5: Seven MCP connectors the author now relies on; triage 2026-07-18

  • clip rated 5: Design-to-code prototype handoff workflow; triage 2026-07-18

  • clip rated 5: Overview of the Fable 5 model; triage 2026-07-18

  • clip rated 5: Anthropic’s classifier quietly reroutes Claude Fable 5 traffic; triage 2026-07-18

  • clip rated 4: A more disciplined AI-assisted dev methodology; triage 2026-07-18

  • clip rated 4: Security researcher’s findings on prompt-injection defenses; triage 2026-07-18

  • clip rated 4: Third-party tool named ‘Graphify’ (unrelated to your own /graphify skill); triage 2026-07-18