skillwiki 0.9.61 → 0.9.63

package/dist/skillwiki-mcp.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 import {
   runSkillwikiMcpStdio
-} from "./chunk-FU462DVS.js";
+} from "./chunk-TUFQZ5K4.js";
 import "./chunk-7I2TPIV5.js";
 
 // src/mcp-entry.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "skillwiki",
-  "version": "0.9.61",
+  "version": "0.9.63",
   "type": "module",
   "bin": {
     "skillwiki": "dist/cli.js",

package/skills/.claude-plugin/plugin.json

@@ -1,8 +1,8 @@
 {
   "name": "skillwiki",
-  "version": "0.9.61",
+  "version": "0.9.63",
   "skills": "./",
-  "description": "Project-aware Karpathy-style knowledge base for Claude Code: 18 prompt-only skills (wiki-*, proj-*, using-skillwiki) backed by the deterministic `skillwiki` CLI.",
+  "description": "Project-aware Karpathy-style knowledge base for Claude Code: 19 prompt-only skills (wiki-*, proj-*, using-skillwiki) backed by the deterministic `skillwiki` CLI.",
   "author": {
     "name": "karlorz",
     "url": "https://github.com/karlorz"

package/skills/.codex-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "skillwiki",
-  "version": "0.9.61",
-  "description": "Project-aware Karpathy-style knowledge base for Codex with 18 prompt-only skills backed by the deterministic skillwiki CLI.",
+  "version": "0.9.63",
+  "description": "Project-aware Karpathy-style knowledge base for Codex with 19 prompt-only skills backed by the deterministic skillwiki CLI.",
   "author": {
     "name": "karlorz",
     "url": "https://github.com/karlorz"
@@ -22,7 +22,7 @@
   "interface": {
     "displayName": "SkillWiki",
     "shortDescription": "Project-aware wiki skills for Codex agents",
-    "longDescription": "18 prompt-only skills (wiki-*, proj-*, using-skillwiki) for deterministic, project-aware knowledge workflows.",
+    "longDescription": "19 prompt-only skills (wiki-*, proj-*, using-skillwiki) for deterministic, project-aware knowledge workflows.",
     "developerName": "karlorz",
     "category": "Productivity",
     "capabilities": [

package/skills/README.md

@@ -5,6 +5,12 @@ or the Claude/Codex/Antigravity plugin packaging paths.
 
 Current package inventory: **18 skills**.
 
+Publication policy: new or updated typed-knowledge and meta pages must use
+`skillwiki page publish` from a temporary draft, inspect its dry-run, and add
+`--write` only after the preview succeeds. Do not directly publish the final
+typed path or separately edit its index/log entries; immutable raw sources and
+non-typed project work retain their existing workflows.
+
 | Namespace | Skills |
 |---|---|
 | `wiki-*` | `wiki-init`, `wiki-ingest`, `wiki-query`, `wiki-lint`, `wiki-crystallize`, `wiki-audit`, `wiki-archive`, `wiki-reingest`, `wiki-adapter-prd`, `wiki-add-task`, `wiki-sync`, `wiki-canvas`, `wiki-gate-plan-mode` |

package/skills/agents/wiki-ingest.md

@@ -6,74 +6,67 @@ color: green
 tools: ["Read", "Write", "Edit", "Bash", "Grep", "Glob"]
 ---
 
-You are a vault ingestion specialist converting source material (URLs, files, text) into typed-knowledge pages with raw provenance. You follow the N6/N7/N8 pipeline: guard → fetch → hash → generate → validate → write. You operate autonomously during maintenance cycles.
+You are a vault ingestion specialist converting source material (URLs and readable local files) into typed-knowledge pages with raw provenance. You follow the N6/N7/N8 pipeline: guard → fetch → hash → generate → validate → write. You operate autonomously during maintenance cycles.
 
 ## When to invoke
 
 - **URL ingestion.** Dev-loop spawns you with URLs to fetch and convert to knowledge pages.
 - **File ingestion.** Local files need to be captured as raw sources and distilled into concept pages.
-- **Batch ingestion.** Multiple sources to process before a single index/log update.
+- **Batch ingestion.** Multiple sources to process through deterministic raw capture and shared publication.
 - **Raw promotion.** A raw/transcripts/ capture is ready for promotion to a typed-knowledge page.
 
 **Your Core Responsibilities:**
 1. Guard: run `skillwiki fetch-guard <url>` for URL sources
-2. Fetch content and write raw file with sha256
-3. Compose typed-knowledge page(s) with citations
-4. Validate every page before writing index/log
-5. Apply writes in order: raw → page(s) → index.md → log.md
+2. Stage pasted text as a readable external file before capture; `skillwiki ingest` does not accept literal text
+3. Capture each source through `skillwiki ingest`
+4. Let the shared publisher own typed-page, taxonomy, index, and structural-log writes
+5. Preserve raw-only recovery state when typed publication fails
 
 **Execution Process:**
 
 1. **Resolve vault and language.** Run `skillwiki path` and `skillwiki lang`.
 2. **Guard (URL sources).** For each URL: `skillwiki fetch-guard <url>`. If non-zero, STOP.
-3. **Fetch.** Fetch content.
-4. **Sensitive content guard.** Before writing or filing any vault page, scan the source and generated body for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact generated prose before writing. If the source itself must remain raw and contains a live secret, STOP instead of preserving it.
-5. **Write raw and hash.** Write raw file at `raw/articles/<slug>.md` with proper frontmatter (`source_url`, `ingested`, `sha256` placeholder). Run `skillwiki hash <raw-file>` and embed the result in `sha256:`.
-6. **Generate page(s).** Compose typed-knowledge pages with:
-   - Proper frontmatter (`title`, `type`, `tags` from SCHEMA.md taxonomy, `provenance`, `sources`)
-   - `## TL;DR` as first section — 1–3 bullet summary
-   - `^[raw/...]` citations for every factual claim
-   - Mermaid diagram if tagged `architecture` or explaining workflows
-   - `confidence: low` if only one source cited
-   - For generated `comparisons/` pages or evaluation-style `queries/` pages, end with:
-     ```markdown
-     ## Decision Closeout
-
-     Disposition: no-op | concept | ADR | work-item | evidence-needed
-     Reason: ...
-     Follow-up: ...
-     ```
-     Use exactly one disposition. This is a prompt convention, not a validator rule.
-7. **Validate.** For each page: `skillwiki validate <page>`. If any non-zero, fix issues and re-validate. Do NOT proceed until all pages pass.
-8. **Apply writes in order:** raw file(s) → page(s) → update `index.md` → append `log.md`.
+3. **Stage pasted text as a file source.** `skillwiki ingest` accepts a readable local source file or HTTP(S) URL; it does **not** accept literal pasted text. For a paste, stage the exact text in a temporary file outside the vault, then pass its path to the normal command:
+   ```bash
+   skillwiki ingest <staged-paste-path> \
+     --vault <resolved-vault> \
+     --type <entity|concept|comparison|query> \
+     --title "<title>" \
+     --tags "<tag1,tag2>" \
+     --provenance <research|project>
+   ```
+   Record the staged path and exact command inputs before execution. If ingestion or typed-page publication fails, retain the staged source and exact command inputs for retry. Remove the staged file only after `skillwiki ingest` exits 0 after typed-page publication; confirm the non-dry-run result reports its raw path, typed path, and publisher operation. Do not stage pasted text inside the vault.
+4. **Sensitive content guard.** Before filing any vault page, scan the source and generated inputs for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact generated prose before writing. If the source itself must remain raw and contains a live secret, STOP instead of preserving it.
+5. **Feature-detect publication.** Run `skillwiki page publish --help`. If unavailable, fail closed and leave typed output unpublished; update the active SkillWiki CLI/plugin channel first.
+6. **Ingest and publish.** Use `skillwiki ingest` for deterministic source capture and typed-page publication. The command writes an immutable raw source first and delegates the typed page, taxonomy, index, and structural log entry to the shared page publisher. Supply the resolved vault, type, title, tags, and provenance through the command options.
+7. **Recovery.** Never create the final typed page or edit index.md/log.md directly. A raw-only result after publication failure is valid recovery state. Keep the exact command inputs and retry; do not delete or overwrite the raw source.
 
 ### Batch mode
 When multiple sources are provided:
-- Execute steps 2–6 per source individually
-- Accumulate all raw files and pages in memory
-- Fail fast: if any page fails validation, STOP and report all failures
-- Deduplicate: check sha256 against existing vault raw sources
-- Single index/log update after ALL sources validate
-- Report progress after each source validates
+- Execute steps 2–7 per source individually, using one `skillwiki ingest` command per source
+- Fail fast: if an ingest command returns nonzero, STOP and report the retained raw-only state, if any, with its exact retry inputs
+- Let `skillwiki ingest` preserve immutable raw capture and skip sources whose content is already present
+- Report progress after each source completes, including the raw path, typed path or recovery state, and publisher operation ID
 
 **Output Format:**
 Return:
 - Sources processed (count)
+- Exact `skillwiki ingest` inputs
 - Raw files written (paths + sha256)
-- Pages generated (paths + types)
-- Validation results
-- Index.md and log.md entries appended
+- Typed pages published (paths + types) or raw-only recovery state
+- Publisher operation IDs and results
 
 **Stop Conditions:**
 - `fetch-guard` non-zero
 - Fetch timeout or size limit exceeded
-- `validate` non-zero on any page (after retry)
+- `skillwiki page publish --help` is unavailable
+- `skillwiki ingest` returns nonzero; retain any raw-only result for retry
 - sha256 already exists in vault (skip, don't duplicate)
 - Source or generated content contains unredacted live credentials or other authenticating secrets
 
 **Forbidden:**
 - Skipping `fetch-guard` for URL sources
-- Updating index/log before all pages validate
+- Creating a final typed page or editing index.md/log.md directly
 - Modifying existing raw files (N9)
 - Writing `[[wikilinks]]` to nonexistent pages — verify first
 - Writing raw ephemeral data to cloud-mounted wiki paths

package/skills/agents/wiki-query.md

@@ -35,7 +35,15 @@ You are a vault search and synthesis specialist using E2 4-signal ranking to fin
 6. **Read top candidates.** Read frontmatter + body of highest-scored pages.
 7. **Synthesize answer.** Compose with explicit citations to candidate pages using `^[page-path]` markers.
 8. **Sensitive content guard.** Before filing a query or comparison page, scan the generated body for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact before writing. If the answer depends on preserving a live secret, STOP and ask for a redacted source or explicit rotation/remediation direction.
-9. **Optional file.** If the task asks to persist: write to `queries/<slug>.md` or `comparisons/<slug>.md` with full frontmatter, validate, then update `index.md` → `log.md`. If the filed page is a research/evaluation answer, recommendation, or comparison, end it with:
+9. **Optional file.** If the task asks to persist, first run `skillwiki page publish --help`. If it is unavailable, fail closed and leave the result unpublished. Otherwise:
+   1. Resolve the vault and create a temporary directory outside the vault.
+   2. Write the complete query/comparison page to `<temp>/page.md`, including final frontmatter, citations, Sources footer, and Decision Closeout.
+   3. Run `skillwiki page publish <temp>/page.md <vault> --target queries/<slug>.md` (or `comparisons/<slug>.md`) and inspect the dry-run.
+   4. Run the identical command with `--write` only when dry-run succeeds.
+   5. On any nonzero result, retain the draft path and STOP; do not edit the final target, index.md, or log.md directly.
+   6. Remove the temporary directory only after the publisher returns complete success.
+
+   If the filed page is a research/evaluation answer, recommendation, or comparison, end it with:
    ```markdown
    ## Decision Closeout
 
@@ -54,15 +62,17 @@ Return:
 - Top candidate pages (ranked, with scores)
 - Synthesized answer with citations
 - Whether result was filed (and path if so)
-- Log entries appended
+- Publisher result (target, operation ID, or retained draft path on failure)
 
 **Stop Conditions:**
 - Zero matching pages found
 - `skillwiki path` returns NO_VAULT_CONFIGURED
 - Generated filed content contains unredacted live credentials or other authenticating secrets
+- `skillwiki page publish --help` is unavailable
 
 **Forbidden:**
-- Filing without `validate` passing
+- Filing without the `skillwiki page publish` dry-run passing
+- Directly creating a final typed page or separately editing `index.md` or `log.md` for its publication
 - Skipping graph refresh when graph.json is missing
 - Accepting wiki claims without filesystem verification
 - Writing live credentials, access keys, tokens, passwords, cookies, bearer headers, private keys, or other authenticating secrets to the vault

package/skills/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@skillwiki/skills",
-  "version": "0.9.61",
+  "version": "0.9.63",
   "private": true,
   "files": [
     "wiki-*",

package/skills/skills/using-skillwiki/SKILL.md

@@ -27,6 +27,7 @@ Invoke a skillwiki skill when the user:
 - Talks about project workspaces, ADRs, or distillation
 - Wants to quickly capture an idea, bug, task, or note without interrupting their workflow
 - Wants to archive or clean up old vault pages
+- Wants to hard-delete a vault path without snapshot resurrection (`wiki-remove`)
 - Needs to detect source drift or re-ingest updated content
 - Has a spec/plan in a non-skillwiki format (CodeStable, RFC, AIDE)
 - Asks about their skillwiki configuration or setup health
@@ -59,6 +60,33 @@ sha256:          # computed by skillwiki hash over body bytes after closing ---
 ## Sensitive Content Policy
 Vault content must not contain live credentials, access keys, tokens, passwords, cookies, bearer headers, private keys, or other authenticating secrets. This includes development-only and local-only credentials. Redact values before filing using `[REDACTED:<kind>]` or `[REDACTED:<kind>:<fingerprint>]`. If a source contains live secrets, stop and ask for a redacted source or explicit rotation/remediation direction; do not preserve the secret in `raw/`.
 
+## Typed-Page Publication Contract
+
+All new or updated typed-knowledge and meta pages MUST be published through
+`skillwiki page publish`. Compose the complete page at an unpublished temporary
+path, run publisher dry-run, then run the same command with `--write`.
+
+- Do not directly create or edit the final typed-page path.
+- Do not directly update `index.md` or append the page's structural log entry.
+- If `skillwiki page publish --help` is unavailable, fail closed and leave the
+  result unpublished; update the active SkillWiki CLI/plugin channel first.
+- `skillwiki validate --apply` is a legacy repair/compatibility path, not the
+  new-page publication path.
+- Non-typed project work items and immutable raw sources keep their existing
+  workflows.
+
+## CLI probe and failsafe (vault-mutating skills)
+
+Vault fleets that combine S3 + GitHub need an explicit delete-intent path. Prefer the skillwiki CLI; when it is missing, agents that still have **git/gh access to the private vault remote** use **FAILSAFE-GIT**.
+
+1. Prefer `skillwiki <subcommand> --help` when the skill requires the CLI.
+2. If skillwiki is missing/unusable but `git` (and optionally `gh`) can reach the private vault remote (`fleet.yaml` `vault_remote` / `origin` → `karlorz/wiki`):
+   - Use **FAILSAFE-GIT** documented in the skill (tombstone under `meta/delete-intents/` + commit + push). Applies to `wiki-remove` and intentional archive/delete paths.
+3. If neither skillwiki nor git/gh private-repo access works: **FAIL CLOSED**.
+   Never bare `rm` / bare `git rm` as a fleet delete (snapshot will resurrect from S3).
+4. Do not auto `npm install -g` skillwiki in headless/goal/satellite sessions.
+5. Delete-intent schema: `vault-delete-intent/v1` JSON under `meta/delete-intents/` (see `wiki-remove`). Git is SSOT for path absence; S3 is a working cache.
+
 ## Portable Source References
 The vault is shared across hosts, so host-local absolute paths are not durable source identity.
 
@@ -84,7 +112,8 @@ The vault is shared across hosts, so host-local absolute paths are not durable s
 | `wiki-lint` | Vault health and lint checks; use `health` for whole-system reports and `lint --summary` for bounded lint buckets |
 | `wiki-crystallize` | Distill the current working session into a typed-knowledge page |
 | `wiki-audit` | Verify raw provenance references and source frontmatter integrity |
-| `wiki-archive` | Archive a typed-knowledge page — move to `_archive/`, remove from index |
+| `wiki-archive` | Archive a typed-knowledge page — move to `_archive/`, remove from index (writes delete-intent) |
+| `wiki-remove` | Hard-delete a path with delete-intent tombstone; failsafe-git if CLI missing |
 | `wiki-reingest` | Detect drift in raw sources (sha256 comparison) and re-ingest updated content |
 | `wiki-add-task` | Quick-capture ideas, bugs, tasks, notes into `raw/transcripts/` without leaving the current workflow |
 | `wiki-adapter-prd` | Map foreign PRD formats (CodeStable, RFC, AIDE, Hermes) into vault pages |
@@ -120,7 +149,7 @@ Use `prd_layer` + `prd_pipeline` from `.claude/dev-loop.config.md` as source of
 
 ## CLI Backbone
 All skills are backed by the `skillwiki` CLI — a deterministic tool with no LLM calls. It handles path resolution, config management, validation, health reporting, and linting. Skills invoke it via Bash for the mechanical parts and use Claude for the creative parts.
-Key CLI subcommands: `init`, `health`, `lint`, `config`, `doctor`, `path`, `lang`, `install`, `fleet context`, `fleet validate`, `graph build`, `archive`, `drift`, `compound`, `tag-sync`, `sync status`, `seed`, `stale`, `observe`, `canvas generate`.
+Key CLI subcommands: `init`, `health`, `lint`, `config`, `doctor`, `path`, `lang`, `install`, `fleet context`, `fleet validate`, `graph build`, `archive`, `remove`, `drift`, `compound`, `tag-sync`, `tag reconcile`, `page publish`, `sync status`, `seed`, `stale`, `observe`, `canvas generate`.
 Run `skillwiki health <vault> --out /tmp/skillwiki-health.json --no-fail` for a bounded whole-system report that includes doctor, lint, vault-sync, query-readiness, source-freshness, risk flags, and self-check coverage. Run `skillwiki lint <vault> --summary` for lint-only bucket counts with capped examples and details commands. Run `skillwiki doctor` to diagnose setup/runtime issues only. Run `skillwiki config list` to see current configuration.
 
 ## Runtime Host Context and Fleet Freshness

package/skills/skills/wiki-archive/SKILL.md

@@ -18,23 +18,36 @@ Run `skillwiki lang` at the start. Generate log entries in the resolved language
 
 Standard four reads (SCHEMA, index, log, project context if applicable).
 
-## Steps
+## Probe
+
+Same matrix as `using-skillwiki` → **CLI probe and failsafe**: PRIMARY (`skillwiki archive`), FAILSAFE-GIT (hand tombstone + `git mv` to `_archive/` + push), or FAIL CLOSED.
+
+## Steps (PRIMARY)
 
 0. Resolve vault: `skillwiki path` and `skillwiki lang`.
 1. Identify the target page. Confirm with the user which page to archive (show full relPath).
-2. Run `skillwiki archive <page> [vault]`. On a vault-sync leaf host where S3 stale originals must be pruned, use `skillwiki archive <page> [vault] --remote seaweed-wiki:cloud/wiki --remote-delete --max-remote-deletes 1` only when that remote path deletion is explicitly intended. Read the JSON output.
+2. Run `skillwiki archive <page> [vault]`. On a vault-sync leaf host where S3 stale originals must be pruned, use `skillwiki archive <page> [vault] --remote seaweed-wiki:cloud/wiki --remote-delete --max-remote-deletes 1` only when that remote path deletion is explicitly intended. Read the JSON output. Successful archive **always** writes a live-path tombstone under `meta/delete-intents/` so snapshot cannot resurrect the live key.
 3. Verify with `skillwiki index-check [vault]` — confirm no ghost entries remain.
 4. Run `skillwiki lint [vault]` — check for broken wikilinks from other pages that still reference the archived page. If found, update those pages to point to the replacement or remove the stale link.
 5. **Raw file archiving (N9 Reingest Protocol only):** When archiving a `raw/` file due to content drift, update ALL `^[raw/...]` citation markers and `sources:` frontmatter entries that reference the old path. Change `raw/articles/foo.md` to `_archive/raw/articles/foo.md` in every referencing page. Verify with `skillwiki audit` that no broken markers remain.
 6. Append a `log.md` entry: `## [{date}] archive | {relPath} → _archive/{subdir}/`.
+7. Commit/push so tombstone + archive land on private `main`.
+
+## FAILSAFE-GIT (no skillwiki)
+
+1. Confirm path. Write `meta/delete-intents/<slug>.json` with `action: "archive"`, `source: "failsafe-git"`, schema `vault-delete-intent/v1`.
+2. `git mv` live path → `_archive/<same relPath>` (create dirs as needed); update `index.md` if typed.
+3. Commit with `Delete-Intent` / `Delete-Source: failsafe-git` trailers; push to private `main`.
+4. Optional single-path `rclone deletefile` for the **live** path if remote configured.
 
 ## Reversibility
 
-Archiving is locally reversible: move the file back from `_archive/` to its original directory and re-add the wikilink entry to `index.md`. If `--remote-delete` was used, the stale active-path object is pruned from the remote after the archive move, but the archived copy remains and a restore will republish the active path on the next push.
+Archiving is locally reversible: move the file back from `_archive/` to its original directory, re-add the wikilink entry to `index.md`, and **delete the matching tombstone** under `meta/delete-intents/`. If `--remote-delete` was used, the stale active-path object is pruned from the remote after the archive move, but the archived copy remains and a restore will republish the active path on the next push once the tombstone is cleared.
 
 ## Stop conditions
 
 - `skillwiki archive` returns non-zero exit code (page not found, already archived, invalid vault).
+- FAIL CLOSED when neither CLI nor private git/gh access is available.
 - User declines to proceed.
 
 ## Forbidden
@@ -42,4 +55,5 @@ Archiving is locally reversible: move the file back from `_archive/` to its orig
 - Archiving `raw/` files outside the N9 Reingest Protocol (raw is immutable except during content-drift reingestion).
 - Archiving raw files without updating all `^[raw/...]` citation markers that reference them.
 - Archiving without user confirmation.
-- Deleting local vault files directly (archive moves locally; remote stale-path pruning is only allowed through the explicit `skillwiki archive --remote ... --remote-delete --max-remote-deletes 1` path).
+- Deleting local vault files with bare `rm` / bare `git rm` without a delete-intent tombstone (causes snapshot resurrection).
+- Remote stale-path pruning only via explicit `skillwiki archive --remote ... --remote-delete` (PRIMARY) or bounded single-path rclone in FAILSAFE-GIT.

package/skills/skills/wiki-ingest/SKILL.md

@@ -14,29 +14,26 @@ Run `skillwiki lang` at the start. Generate page-body prose, narrative sections,
 2. `index.md`
 3. Last 20–30 entries of `log.md`
 4. (Project context only) `projects/{slug}/README.md` and last ~5 work-item logs.
-## Steps (in order — N6, N7, N8)
+## Steps (in order — deterministic raw capture and shared publication)
 0. **Resolve vault and language.** Run `skillwiki path` (fail if NO_VAULT_CONFIGURED) and `skillwiki lang`. Use the resolved vault path for all writes; use the canonical language for all generated prose.
 1. **Guard.** For each URL: run `skillwiki fetch-guard <url>`. If exit ≠ 0, STOP and surface the error. Do not retry.
 2. **Fetch.** Use `web_fetch` (or read local file) under Layer 2 controls (the CLI Layer 2 fetcher applies in tests; in skill runtime use `web_fetch` directly and treat any error as STOP).
    - **Portable local-source rule:** follow `using-skillwiki` → Portable Source References. Do not use `source_url: file:///...` as the canonical durable reference; prefer commit-pinned GitHub `blob/<commit>/<path>` when resolvable, else empty `source_url` plus portable repo-relative prose.
-3. **Identity guard.** Before writing raw files, ensure the target raw filename/title, `source_url`, fetched H1/title, and early body subject agree. If `skillwiki ingest` reports `INGEST_VALIDATION_FAILED` with `source identity conflict`, STOP. Do not fix by renaming after the fact; choose the correct title/source pair or ask the user.
-4. **Sensitive content guard.** Before writing or filing any vault page, scan the source and generated body for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact generated prose before writing. If the source itself must remain raw and contains a live secret, STOP instead of preserving it.
-5. **Hash.** Write the raw file (frontmatter + body). Run `skillwiki hash <raw-file>` and embed the result in raw frontmatter `sha256:`.
-6. **Generate page(s).** Compose typed-knowledge page(s) with citations pre-attached (`^[raw/...]` markers). Every page MUST include:
-- `> **TL;DR:**` blockquote as the first content after the title heading — a one-sentence summary of the page's key takeaway (under 200 chars). See SCHEMA.md `## TL;DR Convention`.
-- For pages tagged `architecture` or explaining workflows/systems: include a Mermaid diagram (`graph TB` or `sequenceDiagram`) in the body. Follow Obsidian-compatible Mermaid rules (see SCHEMA.md `## Mermaid Diagrams`).
-For generated `comparisons/` pages or evaluation-style `queries/` pages, end the body with:
-```markdown
-## Decision Closeout
-
-Disposition: no-op | concept | ADR | work-item | evidence-needed
-Reason: ...
-Follow-up: ...
-```
-Use exactly one disposition. Keep this as a prompt convention, not a validator rule.
-7. **Validate.** For each generated page: run `skillwiki validate <page>`. If exit ≠ 0, STOP — do not write index/log.
-8. **Apply writes in order.** raw → page(s) → `index.md` → `log.md`.
-9. **Confidence flag.** If only one source is cited, set `confidence: low`.
+3. **Stage pasted text as a file source.** `skillwiki ingest` accepts a readable local source file or HTTP(S) URL; it does **not** accept literal pasted text. For a paste, stage the exact text in a temporary file outside the vault, then pass its path to the normal command:
+   ```bash
+   skillwiki ingest <staged-paste-path> \
+     --vault <resolved-vault> \
+     --type <entity|concept|comparison|query> \
+     --title "<title>" \
+     --tags "<tag1,tag2>" \
+     --provenance <research|project>
+   ```
+   Record the staged path and exact command inputs before execution. If ingestion or typed-page publication fails, retain the staged source and exact command inputs for retry. Remove the staged file only after `skillwiki ingest` exits 0 after typed-page publication; confirm the non-dry-run result reports its raw path, typed path, and publisher operation. Do not stage pasted text inside the vault.
+4. **Identity guard.** Before writing raw files, ensure the target raw filename/title, `source_url`, fetched H1/title, and early body subject agree. If `skillwiki ingest` reports `INGEST_VALIDATION_FAILED` with `source identity conflict`, STOP. Do not fix by renaming after the fact; choose the correct title/source pair or ask the user.
+5. **Sensitive content guard.** Before writing or filing any vault page, scan the source and generated body for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact generated prose before writing. If the source itself must remain raw and contains a live secret, STOP instead of preserving it.
+6. **Feature-detect publication.** Run `skillwiki page publish --help`. If unavailable, fail closed and leave typed output unpublished; update the active SkillWiki CLI/plugin channel first.
+7. **Ingest and publish.** Use `skillwiki ingest` for deterministic source capture and typed-page publication. The command writes an immutable raw source first and delegates the typed page, taxonomy, index, and structural log entry to the shared page publisher. Supply the resolved vault, type, title, tags, and provenance through the command options.
+8. **Recovery.** Never create the final typed page or edit index.md/log.md directly. A raw-only result after publication failure is valid recovery state. Keep the exact command inputs and retry; do not delete or overwrite the raw source.
 ## Provenance defaults
 - Default `provenance: research`.
 - If cwd is inside `projects/{slug}/`, set `provenance: project` and add `provenance_projects: ["[[slug]]"]`.
@@ -50,20 +47,19 @@ Raw ephemeral data (market feeds, logs, transient JSON) must be written to the *
 - Fetch timeout / size limit exceeded.
 - `INGEST_VALIDATION_FAILED` with `source identity conflict`.
 - Source or generated content contains unredacted live credentials or other authenticating secrets.
-- `validate` non-zero on any page.
+- `skillwiki page publish --help` is unavailable.
+- `skillwiki ingest` returns nonzero; retain any raw-only result for retry.
 - sha256 already exists in vault for the same source.
 ## Forbidden
 - Skipping `fetch-guard`.
-- Updating `index.md` or `log.md` before all pages validate.
+- Creating a final typed page or editing `index.md`/`log.md` directly.
 - Modifying any existing file in `raw/`.
 - Writing raw ephemeral data directly to cloud-mounted wiki paths (`~/wiki/`).
 - Writing host-local absolute paths as canonical durable source references (see `using-skillwiki` → Portable Source References).
 - Writing `[[wikilinks]]` to pages that don't exist in the vault. Before linking, verify the target exists: check `index.md` or `ls` the target directory. If the target doesn't exist yet, use plain text instead of a wikilink.
 ## Batch Mode
 When the user provides multiple sources (a directory of files, a list of URLs, or a multi-document input):
-1. **Loop per source.** Execute steps 1–7 for each source individually (guard → fetch → identity guard → sensitive content guard → hash → generate → validate).
-2. **Accumulate, don't write yet.** Collect all raw files and pages in memory. Do not write `index.md` or `log.md` until every source has validated.
-3. **Fail fast.** If any page fails validation, STOP. Report all failures. Do not write index/log for any source.
-4. **Deduplication.** Before writing each raw file, check `sha256` against existing vault raw sources. Skip sources whose content is already present.
-5. **Single index/log update.** After all sources validate, write all raw files and pages, then update `index.md` and `log.md` once.
-6. **Progress.** After each source completes validation, report progress (e.g., "Validated 3/10 sources").
+1. **Loop per source.** Execute steps 1–8 for each source individually, using one `skillwiki ingest` command per source.
+2. **Fail fast.** If an ingest command returns nonzero, STOP and report the retained raw-only state, if any, with its exact retry inputs.
+3. **Deduplication.** Let `skillwiki ingest` preserve immutable raw capture and skip sources whose content is already present.
+4. **Progress.** After each source completes, report the raw path, typed path or recovery state, and publisher operation ID.

package/skills/skills/wiki-query/SKILL.md

@@ -23,7 +23,15 @@ Standard four reads (SCHEMA, index, log, project context if applicable).
 5. **Read top candidates** in full (frontmatter + body).
 6. **Synthesize answer** with explicit citations to the candidate pages.
 7. **Sensitive content guard.** Before filing any query or comparison page, scan the generated body for live credentials, access keys, tokens, passwords, cookies, bearer headers, or private keys. Redact before writing. If the answer depends on preserving a live secret, STOP and ask for a redacted source or explicit rotation/remediation direction.
-8. **Optional file.** If user accepts: write to `queries/<slug>.md` or `comparisons/<slug>.md` with full frontmatter, validate, then update `index.md` then `log.md`. If the filed page is a research/evaluation answer, recommendation, or comparison, end it with:
+8. **Optional file.** If the user accepts, first run `skillwiki page publish --help`. If it is unavailable, fail closed and leave the result unpublished. Otherwise:
+   1. Resolve the vault and create a temporary directory outside the vault.
+   2. Write the complete query/comparison page to `<temp>/page.md`, including final frontmatter, citations, Sources footer, and Decision Closeout.
+   3. Run `skillwiki page publish <temp>/page.md <vault> --target queries/<slug>.md` (or `comparisons/<slug>.md`) and inspect the dry-run.
+   4. Run the identical command with `--write` only when dry-run succeeds.
+   5. On any nonzero result, retain the draft path and STOP; do not edit the final target, index.md, or log.md directly.
+   6. Remove the temporary directory only after the publisher returns complete success.
+
+   If the filed page is a research/evaluation answer, recommendation, or comparison, end it with:
 ```markdown
 ## Decision Closeout
 
@@ -41,6 +49,7 @@ Use exactly one disposition. This is a prompt convention only; do not add CLI en
 When a wiki page (especially a work item `tasks.md`) claims that fixes were applied, features were completed, or files were removed — **verify on disk before accepting the claim**. In one incident, a `tasks.md` marked 6 items DONE but 5 were not actually applied: a script claimed "removed" was still 2020 bytes on disk, a crontab claimed "updated to 30min" was still `*/10`, and a build target claimed "verified has consumers" had no web server serving it.
 **Rule**: After reading a work item that declares completion, run at least one verification command per critical claim (check file existence, grep a config, inspect a crontab). Documents can drift from reality — the filesystem is the source of truth.
 ## Forbidden
-- Filing without `validate` passing.
+- Filing without the `skillwiki page publish` dry-run passing.
+- Directly creating a final typed page or separately editing `index.md` or `log.md` for its publication.
 - Skipping the orientation reads even for "quick" queries.
 - Writing live credentials, access keys, tokens, passwords, cookies, bearer headers, private keys, or other authenticating secrets to the vault.

package/skills/skills/wiki-remove/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: wiki-remove
+description: Hard-delete a vault path with a durable delete-intent tombstone so S3+snapshot cannot resurrect it. Prefer skillwiki remove; if CLI missing, failsafe-git via private GitHub.
+---
+
+# wiki-remove
+
+## When This Skill Activates
+
+- User wants to permanently remove a vault path (not archive-for-reference).
+- Cleanup of stale pages that should not return via snapshot resurrection.
+- User says hard delete, remove for real, stop resurrecting.
+
+## Output language
+
+If `skillwiki` is available: run `skillwiki lang` at the start. Otherwise default to English log prose.
+
+## Pre-orientation
+
+Read SCHEMA/index only as needed. Confirm full vault-relative path with the user before delete.
+
+## Probe (required)
+
+1. `command -v skillwiki && skillwiki remove --help`
+2. `git -C <vault> rev-parse --is-inside-work-tree` and `git remote get-url origin` (expect private `karlorz/wiki` / fleet `vault_remote`)
+3. Auth: `gh auth status` **or** `git ls-remote origin HEAD`
+
+| Result | Mode |
+|--------|------|
+| skillwiki OK | **PRIMARY** |
+| skillwiki fail, git/gh + private repo OK | **FAILSAFE-GIT** |
+| neither | **FAIL CLOSED** — stop; never bare `rm` for fleet effect |
+
+Do not auto `npm install -g skillwiki` in headless/goal/satellite sessions.
+
+## PRIMARY steps
+
+1. Resolve vault: `skillwiki path`.
+2. Confirm path with user.
+3. On a vault-sync leaf when S3 prune is intentional:
+   `skillwiki remove <page> [vault] --remote seaweed-wiki:cloud/wiki --remote-delete --max-remote-deletes 1 --reason "<text>"`
+   Otherwise: `skillwiki remove <page> [vault] --reason "<text>"`
+4. Read JSON: expect `tombstone_path` under `meta/delete-intents/`, `removed` set.
+5. Commit + push via normal wiki-sync / git (tombstone + deletion must reach private `main` for durability).
+6. Report MODE=primary, paths, and whether remote-delete ran.
+
+## FAILSAFE-GIT steps (no skillwiki CLI)
+
+Agent **must** have git (and preferably gh) access to private vault remote.
+
+1. Confirm path with user. Set `PATH_REL` (vault-relative, no `..`).
+2. Write tombstone `meta/delete-intents/<slug>.json` where slug is path with `/` → `__`, e.g. `summaries/foo.md` → `summaries__foo.md.json`:
+
+```json
+{
+  "schema": "vault-delete-intent/v1",
+  "path": "PATH_REL",
+  "action": "remove",
+  "created": "ISO-8601-UTC",
+  "host": "SKILLWIKI_HOST_ID-or-unknown",
+  "actor": "agent-failsafe-git",
+  "reason": "user requested remove; skillwiki CLI unavailable",
+  "source": "failsafe-git",
+  "expires": null
+}
+```
+
+3. Delete local file at `PATH_REL` (and index line for `[[slug]]` if typed page).
+4. `git add` tombstone + path change; commit with trailers:
+
+```text
+Delete-Intent: PATH_REL
+Delete-Source: failsafe-git
+```
+
+5. `git pull --rebase origin main` (or merge fallback per vault-sync ADR) then **`git push origin HEAD:main`**. Require successful push before claiming durable success.
+6. Optional: if `rclone` and `WIKI_REMOTE` exist: `rclone deletefile "$WIKI_REMOTE/PATH_REL"` (single path only).
+7. Report `MODE=failsafe-git`, commit SHA, `S3_PRUNE=done|deferred`.
+
+## Stop conditions
+
+- User declines.
+- FAIL CLOSED (no CLI and no private git access).
+- Push to private wiki fails.
+- Path invalid or under `_archive/` without explicit restore/remove policy.
+
+## Forbidden
+
+- Bare `rm` / bare `git rm` without tombstone.
+- Force-push.
+- Unbounded `rclone sync` / mass remote delete.
+- Claiming fleet delete complete without tombstone on `origin/main` (and push success).
+
+## Reversibility
+
+Hard remove is not locally reversible from `_archive`. Restore requires re-authoring content and deleting the tombstone file under `meta/delete-intents/`, then normal push.

package/skills/using-skillwiki/SKILL.md

@@ -27,6 +27,7 @@ Invoke a skillwiki skill when the user:
 - Talks about project workspaces, ADRs, or distillation
 - Wants to quickly capture an idea, bug, task, or note without interrupting their workflow
 - Wants to archive or clean up old vault pages
+- Wants to hard-delete a vault path without snapshot resurrection (`wiki-remove`)
 - Needs to detect source drift or re-ingest updated content
 - Has a spec/plan in a non-skillwiki format (CodeStable, RFC, AIDE)
 - Asks about their skillwiki configuration or setup health
@@ -59,6 +60,33 @@ sha256:          # computed by skillwiki hash over body bytes after closing ---
 ## Sensitive Content Policy
 Vault content must not contain live credentials, access keys, tokens, passwords, cookies, bearer headers, private keys, or other authenticating secrets. This includes development-only and local-only credentials. Redact values before filing using `[REDACTED:<kind>]` or `[REDACTED:<kind>:<fingerprint>]`. If a source contains live secrets, stop and ask for a redacted source or explicit rotation/remediation direction; do not preserve the secret in `raw/`.
 
+## Typed-Page Publication Contract
+
+All new or updated typed-knowledge and meta pages MUST be published through
+`skillwiki page publish`. Compose the complete page at an unpublished temporary
+path, run publisher dry-run, then run the same command with `--write`.
+
+- Do not directly create or edit the final typed-page path.
+- Do not directly update `index.md` or append the page's structural log entry.
+- If `skillwiki page publish --help` is unavailable, fail closed and leave the
+  result unpublished; update the active SkillWiki CLI/plugin channel first.
+- `skillwiki validate --apply` is a legacy repair/compatibility path, not the
+  new-page publication path.
+- Non-typed project work items and immutable raw sources keep their existing
+  workflows.
+
+## CLI probe and failsafe (vault-mutating skills)
+
+Vault fleets that combine S3 + GitHub need an explicit delete-intent path. Prefer the skillwiki CLI; when it is missing, agents that still have **git/gh access to the private vault remote** use **FAILSAFE-GIT**.
+
+1. Prefer `skillwiki <subcommand> --help` when the skill requires the CLI.
+2. If skillwiki is missing/unusable but `git` (and optionally `gh`) can reach the private vault remote (`fleet.yaml` `vault_remote` / `origin` → `karlorz/wiki`):
+   - Use **FAILSAFE-GIT** documented in the skill (tombstone under `meta/delete-intents/` + commit + push). Applies to `wiki-remove` and intentional archive/delete paths.
+3. If neither skillwiki nor git/gh private-repo access works: **FAIL CLOSED**.
+   Never bare `rm` / bare `git rm` as a fleet delete (snapshot will resurrect from S3).
+4. Do not auto `npm install -g` skillwiki in headless/goal/satellite sessions.
+5. Delete-intent schema: `vault-delete-intent/v1` JSON under `meta/delete-intents/` (see `wiki-remove`). Git is SSOT for path absence; S3 is a working cache.
+
 ## Portable Source References
 The vault is shared across hosts, so host-local absolute paths are not durable source identity.
 
@@ -84,7 +112,8 @@ The vault is shared across hosts, so host-local absolute paths are not durable s
 | `wiki-lint` | Vault health and lint checks; use `health` for whole-system reports and `lint --summary` for bounded lint buckets |
 | `wiki-crystallize` | Distill the current working session into a typed-knowledge page |
 | `wiki-audit` | Verify raw provenance references and source frontmatter integrity |
-| `wiki-archive` | Archive a typed-knowledge page — move to `_archive/`, remove from index |
+| `wiki-archive` | Archive a typed-knowledge page — move to `_archive/`, remove from index (writes delete-intent) |
+| `wiki-remove` | Hard-delete a path with delete-intent tombstone; failsafe-git if CLI missing |
 | `wiki-reingest` | Detect drift in raw sources (sha256 comparison) and re-ingest updated content |
 | `wiki-add-task` | Quick-capture ideas, bugs, tasks, notes into `raw/transcripts/` without leaving the current workflow |
 | `wiki-adapter-prd` | Map foreign PRD formats (CodeStable, RFC, AIDE, Hermes) into vault pages |
@@ -120,7 +149,7 @@ Use `prd_layer` + `prd_pipeline` from `.claude/dev-loop.config.md` as source of
 
 ## CLI Backbone
 All skills are backed by the `skillwiki` CLI — a deterministic tool with no LLM calls. It handles path resolution, config management, validation, health reporting, and linting. Skills invoke it via Bash for the mechanical parts and use Claude for the creative parts.
-Key CLI subcommands: `init`, `health`, `lint`, `config`, `doctor`, `path`, `lang`, `install`, `fleet context`, `fleet validate`, `graph build`, `archive`, `drift`, `compound`, `tag-sync`, `sync status`, `seed`, `stale`, `observe`, `canvas generate`.
+Key CLI subcommands: `init`, `health`, `lint`, `config`, `doctor`, `path`, `lang`, `install`, `fleet context`, `fleet validate`, `graph build`, `archive`, `remove`, `drift`, `compound`, `tag-sync`, `tag reconcile`, `page publish`, `sync status`, `seed`, `stale`, `observe`, `canvas generate`.
 Run `skillwiki health <vault> --out /tmp/skillwiki-health.json --no-fail` for a bounded whole-system report that includes doctor, lint, vault-sync, query-readiness, source-freshness, risk flags, and self-check coverage. Run `skillwiki lint <vault> --summary` for lint-only bucket counts with capped examples and details commands. Run `skillwiki doctor` to diagnose setup/runtime issues only. Run `skillwiki config list` to see current configuration.
 
 ## Runtime Host Context and Fleet Freshness