[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79QW5V898C5THZ2KPFXKSN1":3,"skill-same:01M79QW5V898C5THZ2KPFXKSN1":226},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":20,"frontMatter":21,"refName":22,"commitSha":23,"skillDirectory":24,"forksCount":13,"repositoryHtmlUrl":25,"repositoryHomepageUrl":6,"ownerHtmlUrl":26,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":27,"install":28,"resources":41},"01M79QW5V898C5THZ2KPFXKSN1","vault-maintenance",null,"KaedenMonroe\u002Fpkm-agent-os","KaedenMonroe","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F190544785?v=4","01M79QTR659YW26QMXT75206XX","Kaeden Monroe","01M79QTR6W1TK5X9A91W9XTS9F",0,"MIT","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"# vault-maintenance\n\nA periodic-sweep skill for `ObsVault\u002F`. Each maintenance check runs as its own\nscript + AI review pass, producing a human-readable summary in\n`AgentOS\u002Fmaintenance\u002F`. Invoked on demand — no fixed schedule (see\n`02_Design_Specs.md §21` and `AgentOS\u002Fmaintenance\u002Ffuture_steps.md`'s tiering\nand cadence notes).\n\nThis file is the entry point. Steps are documented as their own subsections\nbelow; new steps can be appended without restructuring existing ones.\n\n---\n\n## Step 1: Link Verification\n\n**Script:** `scripts\u002Flink_check.js`\n**Spec:** `02_Design_Specs.md §21`\n\n### What it checks\n\n1. **Broken wikilinks** — `[[links]]` that don't resolve to an existing note.\n2. **Missing required links** — `200_execution\u002F` notes missing a `theory:`\n   link to a `100_concepts\u002F` note (Rule 2), and `100_concepts\u002F` notes of\n   `type: concept` missing a `derived_from` field.\n3. **Candidate links** — pairs of notes with high embedding similarity\n   (via Smart Connections' `.smart-env\u002F` data) that aren't yet linked.\n\n### Running the check\n\nFrom `AgentOS\u002Fskills\u002Fvault-maintenance\u002F`:\n\n```bash\nnode scripts\u002Flink_check.js                  # incremental (default)\nnode scripts\u002Flink_check.js --mode=full      # full O(n²) sweep — \"deep sweep\"\n```\n\n**Incremental (default):** candidate-link discovery (check 3) only\nre-evaluates pairs where at least one note's mtime is newer than\n`state.json`'s `last_run`. Checks 1–2 (broken links, missing required\nlinks) always run over the whole vault — they're linear scans, not\naffected by vault size the way pairwise similarity is.\n\n**Full (`--mode=full`):** candidate-link discovery compares every pair of\nembedded notes, regardless of `last_run` — the \"deep sweep\" that catches\nold↔old pairs that have always existed but were never flagged. Run this\nperiodically (cadence: manual, see `AgentOS\u002Fmaintenance\u002Ffuture_steps.md`'s\ntiering and cadence notes) — e.g. before a vault review session.\n\nIn both modes, pairs listed in `state.json`'s `dismissed_pairs` are never\nre-flagged.\n\nIf `state.json` doesn't exist, every note is treated as \"changed\" — an\nincremental run is then equivalent to a full sweep.\n\nThis writes `link_check_report.json` (gitignored) to the\n`vault-maintenance\u002F` directory, with this shape:\n\n```json\n{\n  \"mode\": \"incremental\",\n  \"generated_at\": \"2026-06-11T18:30:00.000Z\",\n  \"auto_fixes\": [\n    { \"type\": \"broken-link\", \"file\": \"...\", \"old_target\": \"...\", \"new_target\": \"...\", \"raw\": \"...\", \"new_raw\": \"...\" },\n    { \"type\": \"missing-theory-link\", \"file\": \"...\", \"added\": \"[[Concept]]\" }\n  ],\n  \"flagged\": [\n    { \"type\": \"ambiguous-broken-link\", \"file\": \"...\", \"target\": \"...\", \"candidates\": [\"...\", \"...\"] },\n    { \"type\": \"ambiguous-link-collision\", \"file\": \"...\", \"target\": \"...\", \"candidates\": [\"...\", \"...\"] },\n    { \"type\": \"missing-theory-link\", \"file\": \"...\", \"reason\": \"...\" },\n    { \"type\": \"needs-derived-from-review\", \"file\": \"...\" },\n    { \"type\": \"candidate-link\", \"file_a\": \"...\", \"file_b\": \"...\", \"similarity\": 0.81 }\n  ],\n  \"skipped\": [\n    { \"check\": \"candidate-link-discovery\", \"reason\": \".smart-env\u002F not found\" }\n  ]\n}\n```\n\nIf `skipped` includes `candidate-link-discovery`, Smart Connections hasn't\nbeen enabled for `ObsVault\u002F` yet (a one-time manual setup step) — checks 1–2\nstill ran normally.\n\n### Review pass (AI)\n\n1. **Apply each `auto_fixes` entry** via the Edit tool:\n   - `broken-link`: in `file`, replace the literal text `raw` with `new_raw`.\n   - `missing-theory-link`: in `file`'s frontmatter, set\n     `theory: [\"\u003Cadded>\"]` (or append `\u003Cadded>` if `theory:` already has\n     other entries — shouldn't happen for this fix type, since it only\n     fires when `theory:` is empty).\n\n2. **Resolve each `flagged` entry**, per its `type`:\n   - `ambiguous-broken-link` \u002F `ambiguous-link-collision`: read `file` and\n     each note in `candidates`; if one is clearly the intended target, fix\n     the link as in `broken-link` above. If genuinely unclear, leave it and\n     note it for the user in the summary.\n   - `missing-theory-link`: read `file`; if a `100_concepts\u002F` note is the\n     obvious subject, add `theory: [\"[[\u003CConcept>]]\"]`. Otherwise leave it\n     for the user.\n   - `needs-derived-from-review`: apply the `02_Design_Specs.md §10`\n     objective\u002Fsubjective test. If the note is interpretive\u002Fsubjective and\n     a source note can be identified, add `derived_from: [\"[[\u003CSource>]]\"]`.\n     If the note is objective\u002Fcanonical, no action — absence is correct.\n   - `candidate-link`: read both `file_a` and `file_b`. If the relationship\n     is meaningful (not just topically adjacent), add a `[[link]]` between\n     them in whichever note makes more sense contextually. Otherwise drop it.\n\n3. **Write the summary** to\n   `AgentOS\u002Fmaintenance\u002Flink-verification_\u003CYYYY-MM-DD>.md` (sibling to\n   `ObsVault\u002F`, per `02_Design_Specs.md §3`\u002F`§4` — agent output never lives\n   inside the vault). Include: auto-fixes applied, flagged items resolved\n   (and how), and any items left for the user's review.\n\n4. **Update `state.json`**:\n   - Set `last_run` to the report's `generated_at` value.\n   - For each `candidate-link` resolved by **dropping** it (not adding a\n     link), append `[file_a, file_b]` (in the order given by the report) to\n     `dismissed_pairs`, unless an equivalent pair (in either order) is\n     already present.\n   - This step always runs, even if no `candidate-link` items were dropped —\n     `last_run` must advance so the next incremental run doesn't re-scan\n     unchanged files.\n\nPer `02_Design_Specs.md §8` (AI Authorship Boundary), all of the above is\nstructural management — frontmatter, links, tags — which is in scope for AI\nto handle freely. Interpretive prose remains the user's.\n\n---\n\n## Step 2: Card Ideation Pipeline\n\n**Scripts:** `scripts\u002Fflashcard_draft.js` (sweep-side report), `scripts\u002Fflashcard_push.js` (manual push)\n**Spec:** `02_Design_Specs.md §23`\n\nTurns a concept note's *identified* cards (Stage A\u002FB, the user's own work) into\ndrafted Anki card prose (Stage C, AI). Consistent with the AI Authorship\nBoundary (`02_Design_Specs.md §8`): identifying what's worth testing and why\nis germane load (the user's); decomposing that into atomic, bidirectional card\nprose is extraneous load (AI's) — see §8 for the full germane\u002Fextraneous\nframing.\n\n### Stage A — Identification (the user, no artifact)\n\nSelf-interrogation checklist run against a written concept note — table in\n`02_Design_Specs.md §23.2`. Surfaces which of the 7 archetypes (Definition,\nNotation, Theorem\u002FPrinciple, Sequence, Hierarchy, Narrative Sequence, Cached\nInsight) and\u002For which modifier (Attribution, Distinction, Trigger→Concept,\nCase-Approach — `§23.3`) applies.\n\n### Stage B — Card Ideation (the user, permanent artifact)\n\nIdentification results get written into `150_card_ideation\u002F\u003Cconcept-slug>.md`\n(template: `900_system\u002Ftemplates\u002FCard Ideation.md`) — one heading per\narchetype\u002Fmodifier hit, terse, own words:\n\n```markdown\n---\nconcept: \"[[100_concepts\u002FDual Coding Theory|Dual Coding Theory]]\"\nstatus: draft        # draft -> reviewed -> pushed\n---\n\n## Definition\n- verbal vs visual encoding channels — bidirectional\n\n## Attribution\n- Paivio, 1971 — foundational, worth citing (pairs with Definition)\n\n## Cached Insight\n- inferred: extraneous-load automation boundary generalizes to any\n  admin-vs-interpretation split, not just note-taking\n```\n\n**The `concept:` link must keep the `100_concepts\u002F` folder prefix** — an\nideation note shares its concept note's basename by convention (this file\nabove), so a bare `[[Dual Coding Theory]]` link is ambiguous in Obsidian\n(two notes, same basename, different folders) and may resolve to the\nideation note itself rather than the concept note. The alias (`|Dual Coding\nTheory`) keeps the note's rendered\u002Fgraph-view title clean despite the\nqualified target. `flashcard_draft.js`'s resolver strips the prefix before\nmatching against its `100_concepts\u002F`-only basename index, so don't\n\"simplify\" this to a bare basename — that's a regression, not a cleanup.\n\nThis note is **permanent, not a buffer** — it persists after its cards are\npushed, as the standing record of *why* each card exists (unlike\n`000_inbox\u002F` or `review_queue.md`, both cleared out once processed).\n\n**This note's existence is the trigger** — there is no `flashcard_ready`\ncheckbox. A concept is drafted once `150_card_ideation\u002F\u003Cslug>.md` exists and\nits mtime is newer than `state.json`'s `flashcards.last_run`.\n\n### Running the draft (sweep-side)\n\nFrom `AgentOS\u002Fskills\u002Fvault-maintenance\u002F`, always run both commands together\n(never one without the other — the second depends on the first's freshly\nsynced index when the semantic-duplicate check below runs):\n\n```bash\nnode scripts\u002Fanki_similarity_sync.js   # refresh anki_embedding_index.json (§23.7)\nnode scripts\u002Fflashcard_draft.js\n```\n\n`anki_similarity_sync.js` keeps a local embedding index of the **entire**\nexisting Anki collection fresh — incrementally: it re-embeds only notes whose\nAnkiConnect `mod` timestamp changed since the last sync, reusing cached\nvectors for everything else, and prunes notes no longer in the collection.\nLike the rest of this skill, it silently no-ops (leaves the index as-is) if\nAnkiConnect is unreachable — same \"Prerequisite\" pattern as \"Pushing to Anki\"\nbelow. First run downloads the embedding model's weights once (needs\ninternet); every run after that is fully local\u002Foffline. Optional `--dry-run`\nprints stats without writing anything — useful as a first check against a\nreal collection.\n\n`flashcard_draft.js` itself scans `150_card_ideation\u002F` for notes whose mtime is newer than `state.json`'s\n`flashcards.last_run`, resolves each note's `concept:` frontmatter link back\nto its `100_concepts\u002F` note, and writes `flashcard_draft_report.json`\n(gitignored):\n\n```json\n{\n  \"generated_at\": \"2026-07-16T18:30:00.000Z\",\n  \"drafted_notes\": [\n    {\n      \"note\": \"100_concepts\u002FDual Coding Theory.md\",\n      \"ideation_note\": \"150_card_ideation\u002FDual Coding Theory.md\",\n      \"domain_class\": \"other\",\n      \"context_anchor\": \"Dual Coding Theory\",\n      \"interactivity\": \"high\",\n      \"derivation_heavy\": false,\n      \"image_refs\": [],\n      \"sequence\": null,\n      \"declared_sections\": [\"Definition\", \"Attribution\", \"Cached Insight\"],\n      \"unrecognized_headings\": [],\n      \"stage_a_cross_check\": [\n        \"high element-interactivity detected (dense links or comparison framing) but no Distinction\u002FTrigger->Concept\u002FCase-Approach modifier declared — confirm whether one applies (§23.3).\"\n      ]\n    }\n  ],\n  \"flagged\": [\n    { \"type\": \"unresolved-concept-link\", \"note\": \"150_card_ideation\u002FTypo.md\", \"reason\": \"...\" },\n    { \"type\": \"empty-ideation-note\", \"note\": \"150_card_ideation\u002FBlank.md\", \"reason\": \"...\" },\n    { \"type\": \"unrecognized-heading\", \"note\": \"150_card_ideation\u002FLoose.md\", \"headings\": [\"Random Thought\"], \"reason\": \"...\" },\n    { \"type\": \"duplicate-heading\", \"note\": \"150_card_ideation\u002FMessy.md\", \"headings\": [\"Definition\"], \"reason\": \"...\" },\n    { \"type\": \"image-occlusion\", \"note\": \"100_concepts\u002FRankine.md\", \"image_refs\": [\"900_system\u002Fassets\u002Fdiagram.png\"], \"reason\": \"...\" }\n  ],\n  \"skipped\": []\n}\n```\n\nThis is **drafted-state** incremental: an\nideation note is drafted once per edit; `flashcards.last_run` advances only\nafter your review pass, so a note re-edited before review isn't re-drafted.\n\n### Card model\n\nTwo independent dimensions (`§23.2`\u002F`§23.4`), plus modifiers layered on top\n(`§23.3`):\n\n```\nArchetype:  Definition | Notation | Theorem\u002FPrinciple | Sequence\n            | Hierarchy | Narrative Sequence | Cached Insight\nModifier:   Attribution | Distinction | Trigger->Concept | Case-Approach\nFormat:     Basic (Q&A) | Cloze\n```\n\n**The user declares the archetype\u002Fmodifier directly** in the ideation note's\nsection headings — this script no longer infers archetype from a cold\nconcept note (retired: the old `FAVORED_ARCHETYPES` domain-routing table).\nWhat the script *does* still compute, per the user's decision to keep these as\na **secondary assist, not a gate**:\n\n- `domain_class` \u002F `interactivity` \u002F `derivation_heavy` \u002F `sequence` —\n  the same mechanical detectors as before, now run against the **concept**\n  note and folded into `stage_a_cross_check`: advisory notes flagging where\n  the mechanical signal and the user's declared headings disagree (e.g. high\n  interactivity with no Distinction\u002FTrigger→Concept\u002FCase-Approach declared,\n  or a detected ordered list with no Sequence\u002FNarrative Sequence declared).\n  **Never blocks drafting** — read these, and either draft an additional\n  card\u002Fmodifier the cross-check surfaced, or note in the summary why it\n  doesn't apply.\n- `image_refs` — unchanged, still routes to an `image-occlusion` flag\n  (build manually in Anki).\n- `declared_sections` \u002F `unrecognized_headings` — the ideation note's\n  headings, checked against the `§23.2`\u002F`§23.3` vocabulary. An unrecognized\n  heading isn't an error — it might be deliberate free-form scaffolding —\n  but it's flagged so you can confirm it wasn't a typo (e.g. \"Formula\" no\n  longer being separate from \"Theorem\u002FPrinciple\").\n\n`Sequence`\u002F`Narrative Sequence` content-triggered guidance: `length_mode` (default `\"short\"`) means one test-card (recite the\nwhole sequence) + a small cloze set, one element at a time — not the full\nposition\u002Fsuccessor\u002Fpredecessor matrix (`\"long\"` mode, still deferred; see\n`AgentOS\u002Freference\u002Fflashcard_examples\u002Fworked_examples.md`'s Sequence section\nor [eudoxia0\u002Fspaced-repetition-tools](https:\u002F\u002Fgithub.com\u002Feudoxia0\u002Fspaced-repetition-tools)\nif a note ever needs it).\n\n**Worked examples:** before drafting any archetype\u002Fmodifier for the first\ntime in a session, check\n`AgentOS\u002Freference\u002Fflashcard_examples\u002Fworked_examples.md` — anchoring\nexamples for all 7 archetypes plus the 4 modifiers, showing the wrong\n(over-stuffed) version and the corrected breakdown where relevant. Match\ntheir *shape*, never their *content* (Law 7, Source Fidelity).\n\n### The ten Laws (binding when drafting any card)\n\n1. **Context-Anchoring** — every prompt names the parent framework\u002Fdomain of\n   the tested element (use the directive's `context_anchor`); never bare\n   term→definition.\n2. **Minimum Information \u002F Atomicity** — one fact\u002Frelation\u002Fstep per card; no\n   compound questions; no lists unless ≤3 items and the count is the fact.\n3. **Element-Interactivity Routing** — if `interactivity: \"high\"` (and the\n   cross-check flags it), consider routing to the **Distinction** or\n   **Trigger→Concept** modifier rather than a bare base archetype.\n4. **Procedural-Step Fidelity** — **Case-Approach**-modified cards reuse the\n   step names\u002Fordering from the note (or linked `200_execution\u002F` notes)\n   verbatim.\n5. **Supportive-Info Linking** — **Definition\u002FTheorem\u002FPrinciple** cards\n   reference at least one related concept by name in prompt or answer.\n6. **No-Orphan-Recall** — a card shown out of order must still be answerable\n   from its own content.\n7. **Source Fidelity** — paraphrase the ideation note + concept note's\n   prose; introduce no new facts, examples, or framings. Cached Insight cards\n   are the one exception that's still in-bounds: they restructure an\n   inference the user already recorded in the ideation note, never one AI\n   originates itself (`§8`).\n8. **Whole-Task Exclusion** — never draft a card whose answer is a\n   multi-step derivation\u002Fproof\u002Fprocedure. An ideation note with no declared\n   archetype at all is flagged `empty-ideation-note` rather than drafted;\n   a derivation-heavy concept note with *some* declared archetype still\n   drafts those, per the user's own Stage A judgment call.\n9. **Cueing Hygiene** — no answer-leaking: avoid shared word-roots between\n   prompt and answer, no true\u002Ffalse or \"which of the following,\" no\n   implicit-order enumerations.\n10. **Bidirectionality** — for `Definition`, `Notation`, and\n    Distinction-modified cards, draft both directions (term→def, def→term)\n    unless the reverse is trivially inferable from the term's surface form\n    (e.g. \"microglia are a kind of…?\" when \"micro-\" already signals the\n    answer). Default to drafting both; prune the obvious-reverse case with a\n    one-line justification in the maintenance summary. See the Isomers\n    example in `worked_examples.md` for the obviousness threshold in\n    practice.\n\n### Stage C — AI Structuring \u002F Stage D — Gap Flagging (review\u002Fdraft pass)\n\n1. **Read the ideation note** named in each `drafted_notes` directive — not\n   the concept note cold (`§23.1`). The concept note is still read for\n   supporting prose\u002Fcontext, but the ideation note's headings are what\n   determine which cards get drafted.\n2. **Check the matching archetype\u002Fmodifier's shape** against\n   `AgentOS\u002Freference\u002Fflashcard_examples\u002Fworked_examples.md`; write 1-3\n   finished card entries per declared heading into\n   `AgentOS\u002Fskills\u002Fanki-flashcards\u002Freview_queue.md` (append, format below),\n   following the ten Laws. Before writing, re-check against Law 2\n   (Atomicity) and Law 9 (Cueing Hygiene) specifically. **Skip `Longform\n   Candidate` headings** — a routing marker (`§23`), not an archetype\u002F\n   modifier; never draft a card from one. It's Step 3's input instead (see\n   below).\n3. **Anki semantic-duplicate check (`§23.7`), per drafted card, before\n   appending to the queue.** Pipe this note's drafted card(s) as JSON to\n   `node scripts\u002Fanki_similarity_query.js` via stdin\n   (`{\"cards\":[{\"id\":\"...\",\"front\":\"...\",\"back\":\"...\"}]}`) and read the\n   returned shortlist (`{\"results\":[{\"id\":\"...\",\"shortlist\":[...]}]}`) from\n   stdout. This is a cheap, **recall-biased** cosine-similarity pre-filter\n   against the entire existing Anki collection (typically 0-5 candidates per\n   card) — it is not the final verdict. **For each shortlisted candidate,\n   read both the drafted card and the candidate's Front\u002FBack and judge\n   whether they test the same underlying knowledge, even if worded\n   differently, before flagging.** Only write a `> WARNING:` line (format\n   below) for candidates you confirm as true overlap. Drop cosine hits you\n   judge to be false positives silently — don't record them anywhere; the\n   point is reducing the user's review burden, not shifting noise from\n   \"under-checked\" to \"over-checked.\" This is additive to, not a replacement\n   for, `flashcard_push.js`'s existing exact-match\u002Ftag-match guards at push\n   time.\n4. **Work through `stage_a_cross_check`** for the note (Stage D, gap\n   flagging): for each advisory note, either draft the additional\n   card\u002Fmodifier it points at, or record in the maintenance summary why it\n   doesn't apply. Never silently add a card the user didn't identify in\n   Stage A\u002FB, and never silently drop a cross-check note either — every one\n   gets a visible resolution.\n5. **For each `flagged` entry:** append a `## [\u003Cnote>] — FLAGGED` block\n   recording the reason (`unresolved-concept-link`, `empty-ideation-note`,\n   `unrecognized-heading`, `duplicate-heading`, or `image-occlusion`). Never\n   write a card for these. `duplicate-heading` is almost always leftover\n   `Card Ideation.md` template scaffolding pasted alongside real content\n   (e.g. a second empty `## Definition` after the note already has real\n   content under its first `## Definition`) — merge the duplicate's content\n   into the first occurrence (or delete it if empty) rather than just\n   leaving it flagged. Leave the template's reference comment block alone —\n   it's meant to persist as the user's own working reference, not get cleaned\n   up.\n6. **Tag each card entry** with `domain::sub_domain` + a concept-slug\n   (kebab-case of the tested element) **and** an `Ideation:` field pointing\n   back at the ideation note (`§23` — needed for the push-time status flip\n   below).\n7. **Flip the ideation note's frontmatter** `status: draft` → `status:\n   reviewed` (Edit tool) once its cards are written into the queue.\n8. **Advance state:** set `state.json`'s `flashcards.last_run` to the\n   report's `generated_at` value. Always runs, even if zero cards were\n   drafted.\n9. **Summary:** note what was drafted\u002Fflagged\u002Fcross-checked in the\n   maintenance summary (`AgentOS\u002Fmaintenance\u002F`), same as Step 1.\n\nQueue entry format (one per `---`-delimited block):\n\n```markdown\n## [100_concepts\u002FDual Coding Theory.md] — Definition (Basic)\n**Front:** In Dual Coding Theory, what does the Redundancy Effect describe?\n**Back:** ...\n**Tags:** cognitive-science::dual-coding-theory, redundancy-effect\n**Ideation:** 150_card_ideation\u002FDual Coding Theory.md\n---\n```\n\nIf step 3's semantic-duplicate check confirms overlap with an existing Anki\nnote, append a `> WARNING:` line to the entry (same mechanism\n`flashcard_push.js` already uses for its own warnings — `lib\u002Fqueue.js`\nserializes\u002Fstrips it transparently):\n\n```markdown\n## [100_concepts\u002FDual Coding Theory.md] — Definition (Basic)\n**Front:** In Dual Coding Theory, what does the Redundancy Effect describe?\n**Back:** ...\n**Tags:** cognitive-science::dual-coding-theory, redundancy-effect\n**Ideation:** 150_card_ideation\u002FDual Coding Theory.md\n> WARNING: possible duplicate of existing Anki note 1690000000123 (deck: ObsVault) — Front: \"...\" — cosine similarity 0.81 — LLM-confirmed same underlying knowledge; review before pushing.\n---\n```\n\nA `Sequence` note (`length_mode: \"short\"`) yields a small block of its own —\nstill one entry per `---`, archetype labeled `Sequence`:\n\n```markdown\n## [100_concepts\u002FTaxonomic_Rank.md] — Sequence (Basic)\n**Front:** In the taxonomic rank system, recite the four nested ranks from largest to smallest.\n**Back:** Domain, kingdom, phylum, class.\n**Tags:** biology::taxonomic-rank, taxonomic-rank-units\n**Ideation:** 150_card_ideation\u002FTaxonomic_Rank.md\n---\n## [100_concepts\u002FTaxonomic_Rank.md] — Sequence (Cloze)\n**Front:** In the taxonomic rank system, the largest rank is the {{c1::domain}}.\n**Back:** ...\n**Tags:** biology::taxonomic-rank, taxonomic-rank-units\n**Ideation:** 150_card_ideation\u002FTaxonomic_Rank.md\n---\n```\n\n> **Math syntax — MathJax, not LaTeX dollar signs.** Anki renders math via\n> MathJax; it does **not** recognise bare `$...$` \u002F `$$...$$` delimiters by\n> default. Always use:\n> - Inline math: `\\(...\\)` — e.g. `\\(A^k B\\)`, `\\(\\mathscr{C}\\)`\n> - Block\u002Fdisplay math: `\\[...\\]`\n>\n> Pushing cards with `$...$` produces raw-text fields that display as literal\n> dollar-sign strings in Anki (confirmed 2026-06-26).\n\n> **Code syntax — HTML tags, not markdown fences.** `flashcard_push.js`\n> copies Front\u002FBack straight into Anki fields with no markdown-to-HTML\n> conversion (`buildNote()`, `scripts\u002Fflashcard_push.js`), the same reason\n> bare `$...$` fails for math. Markdown code spans\u002Ffences render as literal\n> backtick characters in Anki. Always use:\n> - Inline code: `\u003Ccode>...\u003C\u002Fcode>` — e.g. `\u003Ccode>[x for x in range(10)]\u003C\u002Fcode>`\n> - Multi-line snippets: `\u003Cpre>\u003Ccode>...\u003C\u002Fcode>\u003C\u002Fpre>`\n>\n> Worked examples (this file, `worked_examples.md`) can keep ordinary\n> markdown backticks for readability — only queue entries and pushed fields\n> need the HTML form.\n\n### Anki semantic-duplicate index (`§23.7`)\n\n`anki_similarity_sync.js` and `anki_similarity_query.js` (`scripts\u002Flib\u002Fanki_similarity.js`)\nadd a duplicate-knowledge check that's independent of and earlier than\n`flashcard_push.js`'s existing exact-match\u002Ftag-match guards. It embeds\nFront\u002FBack text with a local model (`TaylorAI\u002Fbge-micro-v2`, via\n`@huggingface\u002Ftransformers` — no API key, no remote calls after the\none-time model download) rather than doing exact-string or exact-tag\nmatching, so it can catch a card that tests the same knowledge as an\nexisting one but is worded differently.\n\n- **`anki_embedding_index.json`** (gitignored, lives in this skill's own\n  directory — never in `ObsVault\u002F`, per Hard Rule 2): `{ schema_version,\n  model, notes: { \"\u003CnoteId>\": { mod, deck, front, back, vec } } }`. Built and\n  incrementally refreshed by `anki_similarity_sync.js` from AnkiConnect's\n  `notesInfo` (per-note `mod` timestamp drives the reuse\u002Fre-embed decision)\n  and `cardsInfo` (deck name — `notesInfo` doesn't return it). These two\n  AnkiConnect actions are new alongside the already-documented\n  `findNotes`\u002F`addNote`.\n- **Threshold** (`ANKI_SIMILARITY_THRESHOLD` in `lib\u002Fanki_similarity.js`,\n  currently `0.70`): deliberately recall-biased — it's a pre-filter, not a\n  verdict, since Stage C step 3 above absorbs the false-positive cost on a\n  small shortlist. This is a reasoned starting guess, not an empirically\n  measured number. After the first few real runs, spot-check shortlist sizes\n  (target ballpark 0-5 per drafted card): lower the constant if a\n  known-near-duplicate isn't surfacing, raise it if shortlists are\n  frequently maxed out with obviously-irrelevant candidates.\n- **Dependency:** `@huggingface\u002Ftransformers` is declared in this skill's\n  `package.json` — run `npm install` in\n  `AgentOS\u002Fskills\u002Fvault-maintenance\u002F` before first use.\n\n### Pushing to Anki (manual, on \"push\")\n\nWhen the user says **\"push\"**, run:\n\n```bash\nnode scripts\u002Fflashcard_push.js\n```\n\n**Prerequisite:** Anki desktop + the AnkiConnect add-on must be running\nlocally (like Step 1's Smart Connections dependency). If unreachable, the push\nreports `skipped` and changes nothing.\n\nIt parses remaining **card** entries (FLAGGED entries are skipped), and applies\nthree duplicate guards: **Layer 0 (generation-phase)** — a card identical\n(same note + Front) to one already staged in this run is a drafting-overlap\nduplicate; it's warned and retained, never pushed twice. **Layer 2** —\n`findNotes tag:\u003Cconcept-slug>`; if matches exist in Anki, warns and leaves the\nentry in place with the warning appended (no add). **Layer 1** — otherwise\n`addNote` to deck `ObsVault` with `allowDuplicate: false`, so Anki's own\nfirst-field guard rejects an exact existing duplicate. Pushed entries are\nremoved from the queue; warned ones stay for the user to resolve.\n\n**Ideation status flip (§23, fully automated per the user's decision):** the\nresult object includes `ideation_status: [{ ideation, status, remaining }]`\n— one entry per ideation note referenced by any card processed this run.\n`status: \"pushed\"` means every one of its cards cleared the queue this run;\n`status: \"partial\"` means some remain (`remaining` count). For every\n`\"pushed\"` entry, **flip that ideation note's frontmatter** `status:\nreviewed` → `status: pushed` (Edit tool) — the script itself never writes\ninto `ObsVault\u002F` (Hard Rule 2); it only computes which notes qualify.\n\n## Step 3: Long-Form Candidate Flagging\n\n**Script:** `scripts\u002Flongform_candidates.js`\n**Spec:** `AgentOS\u002Freference\u002Fplans\u002Fspec_longform-practice_2026-07-17.md`\n(FR-03–05), Phase 3 of the `longform-practice` roadmap.\n\n### What it checks\n\nScans `100_concepts\u002F` and `200_execution\u002F` for the **composition\ndiscriminator** — chains of procedures applied in an applied context (proof\u002F\nderivation\u002Fmulti-step-problem signals), not difficulty or atomicity — that\ndistinguish content unsuited to a flashcard from content that fits Anki fine.\nA note qualifies as a candidate if either:\n\n- `detectDerivationHeavy` (`scripts\u002Flib\u002Fflashcards.js`, already used by Step 2\n  to keep derivations off cards) flags step\u002Fproof-chain signals in its body, or\n- it lives in `200_execution\u002F` at all — that folder is worked derivations by\n  definition (filename convention `\u003Ctheory>__\u003Ccase>.md`), so folder\n  membership alone is a belt-and-suspenders candidate signal, independent of\n  the mechanical heuristic.\n\n### Running the check\n\nFrom `AgentOS\u002Fskills\u002Fvault-maintenance\u002F`:\n\n```bash\nnode scripts\u002Flongform_candidates.js\n```\n\n**Incremental only** (no full-sweep mode, unlike Step 1): only notes whose\nmtime is newer than `state.json`'s `longform_candidates.last_run` are\nconsidered, same mechanism as Step 2's `flashcards.last_run`.\n\nEach candidate is then checked against\n`AgentOS\u002Flongform-practice\u002Fstate\u002Feligibility.json` (FR-05's cache, keyed by\nnote path): if a cached entry's `note_updated_seen` is `>=` the note's current\n`updated` frontmatter value, the note is **not** re-flagged (already reviewed,\nnot stale) — it appears in `skipped` instead. Otherwise it's `flagged`.\n\nWrites `longform_candidates_report.json` (gitignored) to the\n`vault-maintenance\u002F` directory:\n\n```json\n{\n  \"generated_at\": \"2026-07-17T18:30:00.000Z\",\n  \"flagged\": [\n    {\n      \"note\": \"100_concepts\u002FUndetermined Coefficients.md\",\n      \"reason\": \"derivation\u002Fproof step-chain signals detected in note body\",\n      \"signals\": { \"derivation_heavy\": true, \"folder\": \"100_concepts\", \"ideation_declared\": true }\n    }\n  ],\n  \"skipped\": [\n    {\n      \"type\": \"cached-eligibility\",\n      \"note\": \"100_concepts\u002FSome Other Concept.md\",\n      \"reason\": \"reviewed_at 2026-07-11T12:00:00.000Z covers current updated: 2026-07-10\"\n    }\n  ]\n}\n```\n\n**`signals.ideation_declared`** (`100_concepts\u002F` candidates only — see\nbelow): whether the concept's sibling `150_card_ideation\u002F\u003Cslug>.md` already\ndeclares a `## Longform Candidate` section. Computed by checking the\nideation note for that heading (`parseIdeationSections`, `scripts\u002Flib\u002F\nflashcards.js` — the same parser Step 2 uses), content untouched; the\nreview pass reads the actual bullets itself. `200_execution\u002F` notes have no\n`150_card_ideation\u002F` sibling by convention (§23.1 — ideation notes scaffold\nconcept notes only), so this signal is always `false` there and the review\npass falls back to the live prompt below.\n\n### Review pass (AI) — the FR-04 propose\u002Freview exercise\n\nThis step is interactive, not a one-shot fix like Steps 1–2's auto-fixes.\nIt's a declare-then-critique pass, same shape as Step 2's Stage A-D:\nThe user proposes first (in writing, when they've already declared it, or live\notherwise); the AI's job is to critique\u002Frefine that proposal against the\nnote, never to originate it.\n\n1. **For each `flagged` entry, get the user's proposal** — branch on\n   `signals.ideation_declared`:\n   - **`true`:** read the concept's sibling `150_card_ideation\u002F\u003Cslug>.md`\n     directly. Its `## Longform Candidate` bullets *are* the proposal —\n     top-level bullets are problem types, nested bullets are case-branches\n     (template: `900_system\u002Ftemplates\u002FCard Ideation.md`). Skip the live\n     prompt below.\n   - **`false`:** prompt the user to propose the problem types that stem from\n     this concept, **including internal case-branches as explicit\n     sub-items** — e.g. undetermined-coefficients' root-overlap cases (no\n     overlap \u002F single-root overlap \u002F double-root overlap) aren't a separate\n     problem type, they're sub-items of one. This is the only path for\n     `200_execution\u002F` candidates (no ideation-note sibling exists for them)\n     and for `100_concepts\u002F` candidates that haven't declared one yet.\n2. **Review the proposal against the note's own content** for completeness\n   and correctness before eligibility-tracking begins for that concept: does\n   the note's worked example imply a case-branch the proposal omitted? Is\n   every case-branch the note actually demonstrates accounted for? Flag gaps\n   back to the user rather than silently accepting or silently filling them in\n   yourself — FR-04 is a check, not an AI-authored proposal, whether the\n   proposal came from the ideation note or a live answer.\n3. **Once confirmed**, write\u002Fupdate the note's entry in\n   `AgentOS\u002Flongform-practice\u002Fstate\u002Feligibility.json`:\n\n   ```json\n   {\n     \"100_concepts\u002FUndetermined Coefficients.md\": {\n       \"reviewed_at\": \"2026-07-17T18:30:00.000Z\",\n       \"note_updated_seen\": \"2026-07-16\",\n       \"status\": \"confirmed\",\n       \"declared_via\": \"ideation-note\",\n       \"problem_types\": [\n         {\n           \"type\": \"particular solution via undetermined coefficients\",\n           \"cases\": [\"no overlap with homogeneous solution\", \"single root overlap\", \"double root overlap\"]\n         }\n       ]\n     }\n   }\n   ```\n\n   `note_updated_seen` is the note's `updated` frontmatter value **at review\n   time** — this is what FR-05's cache check compares against on the next\n   scan, not the review timestamp itself. `declared_via` is `\"ideation-note\"`\n   or `\"live-prompt\"`, matching step 1's branch — kept for traceability, since\n   provenance (who proposed what, and what the AI critique changed) is the\n   point of a declare-then-critique flow.\n4. **Update `state.json`**: set `longform_candidates.last_run` to the report's\n   `generated_at` value. Always runs, even if zero candidates were flagged\n   this pass — advances the incremental scan the same way Steps 1–2 do.\n5. **Write the summary** to\n   `AgentOS\u002Fmaintenance\u002Flongform-candidates_\u003CYYYY-MM-DD>.md`, same convention\n   as Steps 1–2: which notes were flagged, whether each came from a written\n   declaration or a live prompt, the confirmed problem types\u002Fcase-branches\n   for each, and anything left open for the user.\n\n`eligibility.json` is `longform-practice`'s own persisted cache (not a\n`vault-maintenance` report) — Phase 6 (session generation) reads it directly\nto know which problem types\u002Fcases a concept is confirmed eligible for, once\nits flashcards also clear Step 2's separate maturity gate\n(`AgentOS\u002Flongform-practice\u002Fscripts\u002Fmaturity_gate.js`). The two gates are\nindependent: candidate-flagging tracks *what* to practice, the maturity gate\ntracks *whether it's unlocked yet* — a concept can be a confirmed candidate\nlong before its flashcards mature.\n\n## Future steps\n\nAdditional `vault-maintenance` checks (frontmatter compliance, domain-registry\ndrift, orphan-note detection, etc.) and a fixed run schedule are deferred —\nsee `AgentOS\u002Fmaintenance\u002Ffuture_steps.md`'s candidate-checks table. When\nadded, give each its own `## Step N` section above, following the same\nscript + review-pass shape.\n",{},"master","2b1259f0b2995e1eed2d3c4d3dd9adf0335a85b6","AgentOS\u002Fskills\u002Fvault-maintenance","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os","https:\u002F\u002Fgithub.com\u002FKaedenMonroe","2026-07-27 03:37:51",{"source":7,"skillName":5,"commands":29,"prompt":39,"zipAvailable":40},[30,33,36],{"tool":31,"command":32},"npx","npx skills add KaedenMonroe\u002Fpkm-agent-os --skill vault-maintenance",{"tool":34,"command":35},"bunx","bunx skills add KaedenMonroe\u002Fpkm-agent-os --skill vault-maintenance",{"tool":37,"command":38},"pnpm","pnpm dlx skills add KaedenMonroe\u002Fpkm-agent-os --skill vault-maintenance","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002FSKILL.md",true,[42,48,54,58,64,68,72,76,80,84,88,92,96,100,104,108,112,116,120,124,128,132,136,140,144,148,152,156,160,164,168,171,174,178,182,186,190,194,198,202,206,210,214,218,222],{"path":43,"resourceType":44,"resourceTypeDesc":45,"sizeBytes":46,"sourceHtmlUrl":47},".gitignore","OTHER","其他",123,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002F.gitignore",{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"package-lock.json","REFERENCE","参考文档",38263,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fpackage-lock.json",{"path":55,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":56,"sourceHtmlUrl":57},"package.json",333,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fpackage.json",{"path":59,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":62,"sourceHtmlUrl":63},"scripts\u002Fanki_similarity_query.js","SCRIPT","脚本",1649,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Fanki_similarity_query.js",{"path":65,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":66,"sourceHtmlUrl":67},"scripts\u002Fanki_similarity_sync.js",2016,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Fanki_similarity_sync.js",{"path":69,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":70,"sourceHtmlUrl":71},"scripts\u002Fflashcard_draft.js",6912,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Fflashcard_draft.js",{"path":73,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":74,"sourceHtmlUrl":75},"scripts\u002Fflashcard_push.js",7014,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Fflashcard_push.js",{"path":77,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":78,"sourceHtmlUrl":79},"scripts\u002Flib\u002Fanki_similarity.js",6210,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fanki_similarity.js",{"path":81,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":82,"sourceHtmlUrl":83},"scripts\u002Flib\u002Fanki.js",552,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fanki.js",{"path":85,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":86,"sourceHtmlUrl":87},"scripts\u002Flib\u002Fcandidate_links.js",2506,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fcandidate_links.js",{"path":89,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":90,"sourceHtmlUrl":91},"scripts\u002Flib\u002Fflashcards.js",10290,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fflashcards.js",{"path":93,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":94,"sourceHtmlUrl":95},"scripts\u002Flib\u002Ffrontmatter.js",1516,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Ffrontmatter.js",{"path":97,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":98,"sourceHtmlUrl":99},"scripts\u002Flib\u002Fqueue.js",3838,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fqueue.js",{"path":101,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":102,"sourceHtmlUrl":103},"scripts\u002Flib\u002Frequired_links.js",2465,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Frequired_links.js",{"path":105,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":106,"sourceHtmlUrl":107},"scripts\u002Flib\u002Fstate.js",3501,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fstate.js",{"path":109,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":110,"sourceHtmlUrl":111},"scripts\u002Flib\u002Fvault.js",1383,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fvault.js",{"path":113,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":114,"sourceHtmlUrl":115},"scripts\u002Flib\u002Fwikilinks.js",3476,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flib\u002Fwikilinks.js",{"path":117,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":118,"sourceHtmlUrl":119},"scripts\u002Flink_check.js",2914,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flink_check.js",{"path":121,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":122,"sourceHtmlUrl":123},"scripts\u002Flongform_candidates.js",5510,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Flongform_candidates.js",{"path":125,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":126,"sourceHtmlUrl":127},"scripts\u002Ftests\u002Fanki_similarity_query.test.js",1394,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fanki_similarity_query.test.js",{"path":129,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":130,"sourceHtmlUrl":131},"scripts\u002Ftests\u002Fanki_similarity_sync.test.js",3157,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fanki_similarity_sync.test.js",{"path":133,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":134,"sourceHtmlUrl":135},"scripts\u002Ftests\u002Fanki_similarity.test.js",7347,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fanki_similarity.test.js",{"path":137,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":138,"sourceHtmlUrl":139},"scripts\u002Ftests\u002Fanki.test.js",1515,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fanki.test.js",{"path":141,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":142,"sourceHtmlUrl":143},"scripts\u002Ftests\u002Fcandidate_links.test.js",4338,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fcandidate_links.test.js",{"path":145,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":146,"sourceHtmlUrl":147},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F000_inbox\u002Fnote_with_links.md",85,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F000_inbox\u002Fnote_with_links.md",{"path":149,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":150,"sourceHtmlUrl":151},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F100_concepts\u002FPartial Derivatives.md",131,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F100_concepts\u002FPartial Derivatives.md",{"path":153,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":154,"sourceHtmlUrl":155},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F100_concepts\u002FSeparable Equations.md",137,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F100_concepts\u002FSeparable Equations.md",{"path":157,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":158,"sourceHtmlUrl":159},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F200_execution\u002Forphan_case__weird.md",94,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F200_execution\u002Forphan_case__weird.md",{"path":161,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":162,"sourceHtmlUrl":163},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F200_execution\u002Fseparable_equations__example_one.md",93,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F200_execution\u002Fseparable_equations__example_one.md",{"path":165,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":166,"sourceHtmlUrl":167},"scripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F900_system\u002Ftemplates\u002FCapture - Reading.md",140,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Ffull_vault\u002F900_system\u002Ftemplates\u002FCapture - Reading.md",{"path":169,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":154,"sourceHtmlUrl":170},"scripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F100_concepts\u002FSeparable Equations.md","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F100_concepts\u002FSeparable Equations.md",{"path":172,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":162,"sourceHtmlUrl":173},"scripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F200_execution\u002Fseparable_equations__example_one.md","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F200_execution\u002Fseparable_equations__example_one.md",{"path":175,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":176,"sourceHtmlUrl":177},"scripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F900_system\u002Ftemplates\u002FCapture - Reading.md",234,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Fsample_vault\u002F900_system\u002Ftemplates\u002FCapture - Reading.md",{"path":179,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":180,"sourceHtmlUrl":181},"scripts\u002Ftests\u002Ffixtures\u002Fsmart_env\u002Fmulti\u002Fsample.ajson",315,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffixtures\u002Fsmart_env\u002Fmulti\u002Fsample.ajson",{"path":183,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":184,"sourceHtmlUrl":185},"scripts\u002Ftests\u002Fflashcard_draft.test.js",9229,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fflashcard_draft.test.js",{"path":187,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":188,"sourceHtmlUrl":189},"scripts\u002Ftests\u002Fflashcard_push.test.js",11192,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fflashcard_push.test.js",{"path":191,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":192,"sourceHtmlUrl":193},"scripts\u002Ftests\u002Fflashcards.test.js",11940,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fflashcards.test.js",{"path":195,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":196,"sourceHtmlUrl":197},"scripts\u002Ftests\u002Ffrontmatter.test.js",1863,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Ffrontmatter.test.js",{"path":199,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":200,"sourceHtmlUrl":201},"scripts\u002Ftests\u002Flink_check.test.js",5513,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Flink_check.test.js",{"path":203,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":204,"sourceHtmlUrl":205},"scripts\u002Ftests\u002Flongform_candidates.test.js",7407,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Flongform_candidates.test.js",{"path":207,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":208,"sourceHtmlUrl":209},"scripts\u002Ftests\u002Fqueue.test.js",4627,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fqueue.test.js",{"path":211,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":212,"sourceHtmlUrl":213},"scripts\u002Ftests\u002Frequired_links.test.js",3028,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Frequired_links.test.js",{"path":215,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":216,"sourceHtmlUrl":217},"scripts\u002Ftests\u002Fstate.test.js",5408,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fstate.test.js",{"path":219,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":220,"sourceHtmlUrl":221},"scripts\u002Ftests\u002Fvault.test.js",2110,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fvault.test.js",{"path":223,"resourceType":60,"resourceTypeDesc":61,"sizeBytes":224,"sourceHtmlUrl":225},"scripts\u002Ftests\u002Fwikilinks.test.js",4466,"https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvault-maintenance\u002Fscripts\u002Ftests\u002Fwikilinks.test.js",[227,233,239,245,251,258],{"skillNo":228,"name":229,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":230,"categories":231,"keywords":232,"publishedAt":18},"01M79QWAPGV4MWQXZD2DSXB3DY","vocab-anki","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fvocab-anki\u002FSKILL.md",[],[],{"skillNo":234,"name":235,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":236,"categories":237,"keywords":238,"publishedAt":18},"01M79QW2K0FFQTX8RW7GT15KZY","transcript-distillation","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Ftranscript-distillation\u002FSKILL.md",[],[],{"skillNo":240,"name":241,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":242,"categories":243,"keywords":244,"publishedAt":18},"01M79QVZXJ69XH1ZNQJ59VHD30","daily-log","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fdaily-log\u002FSKILL.md",[],[],{"skillNo":246,"name":247,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":248,"categories":249,"keywords":250,"publishedAt":18},"01M79QVWWEHN5BZ963895VRR8M","capture-promotion","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Fskills\u002Fcapture-promotion\u002FSKILL.md",[],[],{"skillNo":252,"name":253,"nameZh":6,"description":254,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":255,"categories":256,"keywords":257,"publishedAt":18},"01M79QVSJXDQYXSGBQE6MSH469","review","Clear due memory reviews with free recall — the two-minute habit that makes learning permanent. Use when reviews are due, or the user wants to review, practice, or \"do my engram reviews\".","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Flongform-practice\u002Fengram\u002Fskills\u002Freview\u002FSKILL.md",[],[],{"skillNo":259,"name":260,"nameZh":6,"description":261,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":262,"categories":263,"keywords":264,"publishedAt":18},"01M79QVM3KNV7CNFFYC0X4TAZ2","learn","Learn any topic properly — first-principles curriculum, generation-first tutoring, verified free recall, FSRS scheduling. Use when the user wants to learn, understand, study, or continue studying something.","https:\u002F\u002Fgithub.com\u002FKaedenMonroe\u002Fpkm-agent-os\u002Fblob\u002Fmaster\u002FAgentOS\u002Flongform-practice\u002Fengram\u002Fskills\u002Flearn\u002FSKILL.md",[],[]]