Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →One-screen quick reference for the Origin plugin. Lists the daily verbs, the daily flow, where data lives, and how to view it without a GUI. Use when the user says "help", "what can I do", "list origin commands", "how do I use origin", or invokes `/help`.
.claude/skills/davepoon-help/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 9 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 5% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 1% | 0% |
Print the Origin plugin reference card. Read-only — never calls a tool.
When triggered, output the block below verbatim. No editing, no abbreviating, no embellishing. The user is asking for the menu.
Origin plugin — daily verbs
/init set up Origin (auto-installs daemon + local memory)
/brief load identity + topic context (start of session)
/capture <x> save one durable memory in flow
/recall <q> search local memory
/distill [t] synthesize pages from clusters (scoped to current repo)
/read <p> preview a distilled page inline
/review <surface> deep audit (surface = captures|revisions); /brief handles daily
/forget <id> delete a memory by ID
/handoff end-of-session ritual (session log + captures)
/debrief alias for /handoff (brief/debrief symmetry)
/help this card
Daily flow (~1 min overhead per session):
1. start session → hook auto-checks daemon, silent if up
2. /brief → ~5 s, load context
3. work normally → Claude proactively /captures durable facts
4. /recall X → as needed for lookups
5. /handoff → ~30 s, narrative session log + captures
Where your data lives (everything under ~/.origin/):
~/.origin/pages/ wiki pages distilled from your memories (md)
~/.origin/sessions/ session logs by date (md)
~/.origin/sessions/_status/ current per-project goals + last-handoff
~/.origin/db/ memories + knowledge graph (symlink to libSQL)
~/.origin/bin/ installed binaries
View it without a GUI:
open ~/.origin/ browse in Finder
code ~/.origin/ open in VS Code
git -C ~/.origin log --oneline timeline of every memory + distill pass
ln -s ~/.origin/pages ~/Vault/origin # symlink into Obsidian for graph view
~/.origin/ is a git repo. Skills auto-commit per logical batch (one per
session, distill pass, or forget). Use git log / git diff / git revert
as a free audit trail. No remote — purely local history.
Three classes of artifact:
- memories: granular, queryable, live in DB only (confirmed = stays in DB)
- pages: synthesized wikis, DB + ~/.origin/pages/*.md projection
- sessions: chronological narrative, ~/.origin/sessions/*.md only
Daemon must run at 127.0.0.1:7878. Hook prints "/origin:init" if down.
Optional upgrades for richer distill cycles:
origin model install local Qwen, no API cost
origin key set anthropic Anthropic API, higher quality/help.does this plugin work", "remind me what verbs are available".
/init success too./recall./init (it diagnoses + auto-installs).| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 9,780 | 3,250 | -67% | 1 | 1 | 0% | 1,593 | 1,506 | -5% | 0 | 0 | — |
case-02 | fail→pass | 16,775 | 3,541 | -79% | 1 | 1 | 0% | 3,003 | 1,595 | -47% | 0 | 0 | — |
case-03 | fail→pass | 11,872 | 5,213 | -56% | 1 | 1 | 0% | 2,028 | 2,130 | +5% | 0 | 0 | — |
case-04 | fail→pass | 4,753 | 2,929 | -38% | 1 | 1 | 0% | 864 | 1,460 | +69% | 0 | 0 | — |
case-05 | fail→pass | 11,904 | 5,586 | -53% | 1 | 1 | 0% | 2,152 | 2,181 | +1% | 0 | 0 | — |
case-06 | pass→pass | 9,659 | 3,244 | -66% | 1 | 1 | 0% | 1,839 | 1,567 | -15% | 0 | 0 | — |
case-07 | fail→pass | 2,957 | 3,072 | +4% | 1 | 1 | 0% | 453 | 1,539 | +240% | 0 | 0 | — |
case-08 | fail→pass | 8,889 | 5,427 | -39% | 1 | 1 | 0% | 1,629 | 2,200 | +35% | 0 | 0 | — |
case-09 | fail→pass | 3,813 | 3,886 | +2% | 1 | 1 | 0% | 509 | 1,371 | +169% | 0 | 0 | — |
case-10 | fail→pass | 9,333 | 4,388 | -53% | 1 | 1 | 0% | 1,653 | 1,445 | -13% | 0 | 0 | — |
case-11 | fail→pass | 9,960 | 4,541 | -54% | 1 | 1 | 0% | 1,774 | 1,538 | -13% | 0 | 0 | — |
case-12 | pass→pass | 7,741 | 2,232 | -71% | 1 | 1 | 0% | 1,556 | 1,212 | -22% | 0 | 0 | — |
case-13 | pass→pass | 10,842 | 3,689 | -66% | 1 | 1 | 0% | 1,915 | 1,446 | -24% | 0 | 0 | — |
case-14 | fail→pass | 7,822 | 1,383 | -82% | 1 | 1 | 0% | 1,483 | 958 | -35% | 0 | 0 | — |
case-15 | fail→fail | 18,171 | 3,255 | -82% | 1 | 1 | 0% | 3,035 | 1,304 | -57% | 0 | 0 | — |
case-16 | fail→pass | 8,111 | 1,309 | -84% | 1 | 1 | 0% | 1,400 | 908 | -35% | 0 | 0 | — |
case-17 | fail→pass | 8,268 | 1,484 | -82% | 1 | 1 | 0% | 1,333 | 975 | -27% | 0 | 0 | — |
case-18 | fail→fail | 12,529 | 2,928 | -77% | 1 | 1 | 0% | 2,151 | 1,236 | -43% | 0 | 0 | — |
case-19 | pass→pass | 10,596 | 1,686 | -84% | 1 | 1 | 0% | 1,917 | 1,046 | -45% | 0 | 0 | — |
case-20 | fail→pass | 9,070 | 1,521 | -83% | 1 | 1 | 0% | 1,432 | 1,016 | -29% | 0 | 0 | — |
case-21 | fail→pass | 9,670 | 1,241 | -87% | 1 | 1 | 0% | 1,605 | 883 | -45% | 0 | 0 | — |
case-22 | fail→fail | 11,422 | 1,558 | -86% | 1 | 1 | 0% | 1,829 | 932 | -49% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +68 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.