History

Absorb NLE, JobNavigator, Ralph, Deterministic-Tool/Agent Partition, and Keelhaven blips Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QgvcBNceJCdTUtvQsvfDb7

dennislwm committed Sep 14, 2026

Update Floci/MiniStack service-depth research and add terraform-provider-aws blip Floci grew from 68 to 150+ services with vendor-documented per-feature depth; MiniStack's landing-page "Full" claims flagged as unverified with several explicit stubs. Position adjusted: Floci stays inner, MiniStack moves center to outer. Added terraform-provider-aws (Trial) citing confirmed first-person use in awscode (2021, real AWS) and a LocalStack runbook in examprep-terraform (2024). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JfnJDR2PfJBV39mSbVwBQ5

dennislwm committed Sep 6, 2026

Absorb archify, LosslessCut, vault-cortex, multi-worker-pm-skill, AIUP, and Local LLM Memory Sizing blips Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JfnJDR2PfJBV39mSbVwBQ5

dennislwm committed Sep 6, 2026

Absorb OpenExecutive and Concat (formerly WolfCut) blips Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018H4U8tRpri1WD1v1ycbPqo

dennislwm committed Sep 4, 2026

Absorb Beads, hoop, mex, Mimik, and Johnny.Decimal MCP server blips Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Aug 28, 2026

Absorb Swamp, nobuzz, Saggar, cartcut, Loadout blips and a token-spend technique; fix stale Herdr repo link Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Aug 27, 2026

Add Repowise, Foreman, GQM/audit/ledger technique blips; fix stale CDN cache bug in fetch_bookmarks.py Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Aug 19, 2026

Absorb aws-cli and litellm blips (Tools/Trial) Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014HBECSVN7FmByEKjs6QDc3

dennislwm committed Aug 17, 2026

Absorb TDD-in-Agent-Loop, faaah, MCP python-sdk, diagram-design blips Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014HBECSVN7FmByEKjs6QDc3

dennislwm committed Aug 17, 2026

Add 18 new radar entries across platforms, techniques, and tools Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014HBECSVN7FmByEKjs6QDc3

dennislwm committed Aug 17, 2026

Add OpenWorker, Palmier Pro, Berthly, and Datasette Plugin Authoring blips Absorbed from bookmarks batch and follow-up research; cursor advanced and _radar.md regenerated. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0146MCncbncPDhFTiryrLi66

dennislwm committed Jul 28, 2026

Add update_cursor.py helper, ClaudeObsidian/HarnessEngineering/LoopEngineering articles, document TrmnlTelegram hosting Cursor writes now go through helper/update_cursor.py instead of an inline heredoc, with settings.json and CLAUDE.md updated to match. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Y5Cu5eydZZcQHBWbAwhgFx

dennislwm committed Jul 23, 2026

Add Apple Guideline-Driven Framework Selection and Datasette Live-Plus-Static Hosting blips Two Techniques blips: comparing iOS/web app-building approaches through Apple App Store review guidelines, and hosting options for a live-plus-static Datasette setup through each platform's own published limits. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Y5Cu5eydZZcQHBWbAwhgFx

dennislwm committed Jul 23, 2026

Retire FireRed-OpenStoryline (disqualified); pivot pipeline editing stage First-person trial in pynanoclip disqualified FireRed-OpenStoryline: the agent has no working hand-off for structured discovery data (routes prompts into speech_rough_cut, empty timeline, failed render), no prompt fix, pre-cut workaround 2/3, vendor patches don't survive git pull. Deterministic recipe-to-CapCut (capcut-cli, playcapcut) replaces it. - Retire FireRed-OpenStoryline: retired/disqualified/outer; rewrite assessment - Personal Video Processing Pipeline: Stage 2 editing pivoted agentic -> deterministic; ring unchanged (no full first-person production run yet) - Fix capcut-cli wikilinks; regenerate _radar.md Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TgMT8xkUvAjP88mxKe7B1d

dennislwm committed Jul 20, 2026

fix: correct 12 mis-ringed blips, retire erm, add cascade review state - Retire erm (disqualified): poor-quality output on real recordings - Demote 12 blips Trial/Adopt -> Assess, ringed on third-party maturity not first-person use (Starlette, ultralytics, Syncthing, Bruno, OpenTelemetry Collector, SLOs, Three-Pass Paper Reading, Twenty Years of SRE, OTel Git Performance, Terraform Testing, k3s on Oracle, C4/Structurizr) - CLAUDE.md: extend first-person gate to Adopt; add Assess-vs-Trial/Adopt disambiguation to stop absorbing maturity as adoption - Add helper/mark_cascaded.py: cascade review state on latest history entry (--pending lists candidates, flag skips unchanged blips); wire into /wiki cascade - Regenerate _radar.md Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TgMT8xkUvAjP88mxKe7B1d

dennislwm committed Jul 20, 2026

feat: add OpenCLI, Harness Handbook, i-have-adhd blips; document trmnl-telegram hosting - Absorb OpenCLI and Harness Handbook (Techniques/Assess/outer) - Absorb i-have-adhd Claude Code plugin (Tools/Assess/inner) - Add trmnl-telegram Hosting Requirement section (always-on daemon) - Pin helper scripts individually in .claude/settings.json; drop broad wildcard - Advance bookmark cursor; regenerate _radar.md (312 blips) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TgMT8xkUvAjP88mxKe7B1d

dennislwm committed Jul 20, 2026

feat: add AI Code Archaeology, DSL-Constrained LLM, Human-First CLI, trmnl-telegram, Conductor blips Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

dennislwm committed Jul 18, 2026

docs: record first-person validation of capcut-cli in article Update CapcutCli.md after installing v0.13.2 and exercising it on macOS this session: doctor passes, read-only inspection parsed real CapCut 3.3.0 drafts, and a mutation-and-undo cycle (add-text then restore) validated the .bak + snapshot safety net. Fixes two wrong command names in the Cross-Platform Note (list-projects/get-project -> projects/info). Adds a Limits section (no final render, translate/caption external deps, JianYing 6.0+ encryption unsupported). Ring unchanged at Assess/inner: no real project rendered to a deliverable yet, so the Trial gate now narrows to one of the user's own drafts edited and exported. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EFBjtPe5mV7ZtS5HJ2Dsza

dennislwm committed Jul 16, 2026

feat: add capcut-cli blip (Tools/Assess/inner) Independent MIT-licensed CLI that edits CapCut and JianYing projects by reading and writing their local JSON draft files directly, with no official API: timeline repair, word-level subtitle export, scene-cut detection, text-style presets, and a stdin queue runner for LLM-driven automation. Placed at inner on source-level vetting (v0.13.2, 170 stars) rather than on adoption: zero runtime dependencies, no install scripts, no dynamic code execution, no network egress on normal edits, and a pre-edit snapshot for undo. Strongest of four CapCut automation tools evaluated and the only one simultaneously licensed, dependency-free, network-quiet by default, and cross-platform. Assess not Trial: no first-person production use yet. A concrete instance of MCP Server Supply-Chain Vetting: among four similarly-named repos, star count did not predict trustworthiness, and only source review separated this from stale unlicensed forks and a networked service binding 0.0.0.0 with no auth. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EFBjtPe5mV7ZtS5HJ2Dsza

dennislwm committed Jul 16, 2026

feat: add erm, Jailbreaker, MCP Server Supply-Chain Vetting blips; add floci-dash to Floci New blips: - erm (Tools/Assess/outer): CLI stripping disfluencies from speech recordings - Jailbreaker (Tools/Assess/center): SpecterOps LLM jailbreak testing platform - MCP Server Supply-Chain Vetting (Techniques/Assess/inner): treat third-party MCP servers as untrusted supply chain rather than trusted plugins The MCP vetting technique is sourced from Canopii's State of MCP Security 2026, which scored 11,524 registry servers: 232 with confirmed dangerous code sinks, 78% with unpinned dependencies, 184 post-publication tool-definition rug pulls, and 31% of auth-declaring endpoints not enforcing it. Servers with 1,000+ stars are over 5x likelier to be high-risk, so popularity inverts as a safety signal. Assess rather than Trial: the gates have no first-person application yet. Also enrich Floci with a Companion Tooling section covering floci-dash, and advance the bookmark cursor to 2026-07-14T12:39:49Z. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EFBjtPe5mV7ZtS5HJ2Dsza

dennislwm committed Jul 16, 2026

docs: condense CLAUDE.md without changing any rule Merge duplicated sections rather than dropping guidance: - Fold "Bookmarks source" into the bookmarks workflow section - Hoist the inline-Python prohibition into one "Helper scripts" section - Flatten the quadrant table; merge the Trial gate into the Trial definition - Absorb the /lens + /wiki section into "External tool research" Drop the /wiki radar ASCII example and symbol legend. Those rules live in helper/radar_diagram.py (POSITION_SYMBOL, RING_ORDER, retired filtering), which is canonical per the helper-scripts rule; the prose was duplication that could only drift out of sync. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EFBjtPe5mV7ZtS5HJ2Dsza

dennislwm committed Jul 16, 2026

feat: add AWS VPC Lattice, Rayfish, AWS Best Practices Skill blips; enrich Floci and Fowler memo series New blips: AWS VPC Lattice (Platforms/Assess/inner, includes CloudFormation reference and hardening guide), Rayfish (Platforms/Assess/center, P2P mesh VPN), AWS Best Practices Skill (Tools/Assess/inner, Claude Code/Codex skill). Enriched Floci with current service count and confirmed no VPC Lattice support; appended a new memo to the Exploring Generative AI series on local model coding viability. Also commits prior pending herdr and OfficeCLI blips already referenced in _blips.json. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Jul 11, 2026

feat: rename /wiki scaffold to /wiki ladder, add /wiki ladder <path> audit command Points wiki-scaffolding at the canonical template in 02claude-code instead of a drifted downstream copy, and adds a read-audit-then-extend command for existing wikis' CLAUDE.md (never bulk-creates Requirements/Decisions/etc, those come on-demand via add req/add test or the artifact-placement ladder). Also updates the Geospatial MCP for AI Planning blip with findings from a real validation attempt: free geocoding gaps, a public routing demo silently substituting driving speed for walking requests, inconsistent keyed-API access across client environments, and a working lookup that still understated an actual walk by half. Ring stays Assess. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Xw7iUGSGjCFqBMj1G1Nthb

dennislwm committed Jul 6, 2026

fix: /wiki scaffold uses pynanoclip template, adds create subcommand step Pylabel-based template lacked the create subcommand pynanoclip's wiki gained after scaffolding; switch the template source and document the substitution so future scaffolds include it and advertise it correctly. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Jul 4, 2026

fix: correct /wiki scaffold prefix handling in GitHub repo links The numeric folder prefix (13, 20) is a local naming convention only and never appears in the GitHub repo name -- confirmed against 13pylabel's own remote. The spec previously baked the prefix into the sibling repo URL, which would have produced a broken link. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Jul 4, 2026

feat: add Personal Video Processing Pipeline blip; add /wiki scaffold subcommand Consolidates the ad-hoc video pipeline description out of CLAUDE.md into a proper Techniques blip documenting considered/rejected paths and per-path limitations. Adds a /wiki scaffold subcommand for graduating a blip into a sibling implementation project wiki. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

dennislwm committed Jul 4, 2026

feat: add 5 blips (Geospatial MCP, QR Label Pipeline, scaffold-docs-skill, TREK, hasp); update article format rules in CLAUDE.md Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

dennislwm committed Jun 26, 2026

feat: add 4 blips from bookmarks batch (2026-06-24) - Prompt Injection as Role Confusion (Techniques/Assess): ICML 2026 framework; CoT Forgery attack at 60% success rate - AgentsView (Tools/Assess): local-first SQLite analytics for Claude Code and 20+ coding agents - Claude Code Undocumented Configuration (Techniques/Assess): hook response fields, skill frontmatter extensions, agent config - Karpathy LLM Coding Guidelines (Techniques/Assess): 4-principle CLAUDE.md behavioral guardrail, 181k stars Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

dennislwm committed Jun 25, 2026

feat: retire Varlock; fix query_blips keyword search; ponytail-audit helper scripts - Retire Varlock blip (superseded by lpassrc); update article ring and Radar Assessment - lpassrc: add cascade history entry confirming sole active env injection role - query_blips.py: fix --keyword to use any() instead of all() (OR not AND) - add_blip.py, update_blip.py: use json.load() instead of json.loads(open().read()) - radar_diagram.py: remove empty f-string, replace used_rings set with any() checks - generate_radar_md.py: fix double-blank-lines; use \n\n join; consolidate retired rows - Regenerate _radar.md (277 active blips) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

dennislwm committed Jun 25, 2026

docs: add article ring sync rule and ponytail section to CLAUDE.md Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

dennislwm committed Jun 24, 2026