@c4a/context-cli 0.7.14 → 0.7.16

package/providers/context/resources/manuals/guides/package-outputs.md

@@ -99,8 +99,14 @@ The website uses a full-width VitePress theme with system fonts, compact navigat
 a wide reading area and a smaller article outline. It starts in
 light mode regardless of the operating system; an explicit reader choice is
 remembered in browser storage. Search runs locally, without a search service.
-Mermaid renders in the browser with a restrained theme; invalid diagrams retain
-their source instead of blocking publication. Code samples and raw HTML are
+Mermaid renders in the browser using the website theme, with fit, original-size,
+zoom, full-screen and source controls. Controls appear on hover or keyboard focus
+(and remain visible on touch devices). Layout and source/copy actions are on the
+left, sizing actions on the right. Source view replaces the diagram. Flowcharts default to ELK and can switch to Dagre without changing the article. Offline Review reports use the same
+viewer and include its resources only when a candidate contains Mermaid.
+Article provenance initially shows ten sources; Show more reveals the full list.
+Invalid diagrams and diagrams with embedded image/HTML resources retain their
+source with a rendering notice instead of blocking publication. Code samples and raw HTML are
 displayed as content, not executed as Vue components or scripts.
 
 The accepted `src/knowledge-map.yaml` controls sidebar organization. Entries
@@ -507,3 +513,31 @@ result instead of guessing a hostname. No new workspace configuration is needed.
 The address is an access hint, not proof of publication or content freshness.
 Query Skills use it only to append related links in the final summary, with no
 network probes or version checks. Packages without the map work as before.
+
+### Shared color theme
+
+Website builds create `src/site/theme.json` with the current default light and
+ dark palettes if the file is missing. Edit this optional JSON file to customize
+all websites and offline Review reports in the workspace. It is configuration,
+not a generated HTML file: builds never overwrite existing values. Deleting it
+restores built-in defaults; Review does not need the file or an existing website
+build. The next website build recreates it.
+
+Each `light`/`dark` object supports `brand`, `accent`, `background`, `surface`,
+`text`, `mutedText`, and `border`. Values are hexadecimal CSS colors (`#RGB`,
+`#RRGGBB`, or `#RRGGBBAA`). Partial objects inherit missing default colors. Invalid
+JSON or colors produce a configuration error identifying the optional file.
+
+A package may override these colors through `kbPackage({ site: { theme: {
+light: { brand: "#6750a4" }, dark: { brand: "#d0bcff" }
+} } })`. Precedence is built-in defaults, workspace file, then `site.theme`.
+A Review with one website package uses that package's overrides. With multiple
+website packages it uses the shared workspace theme, without selecting an
+arbitrary package. New/Modify/approve/reject status colors remain independent.
+
+The compiler projects tokens into VitePress and standalone Review CSS. Mermaid
+reads the same resolved colors. Changes to the theme invalidate website build
+receipts. Arbitrary extension CSS is not copied into Review; custom components
+can consume `--context-brand`, `--context-accent`, `--context-background`,
+`--context-surface`, `--context-text`, `--context-muted-text`, and
+`--context-border` rather than hard-coding colors.

package/providers/context/resources/procedures/document-capture.md

@@ -62,3 +62,10 @@ are available while another capture remains pending. Keep unresolved sources in
 `pending_scopes`; do not cite their unavailable snapshots. The Route returns to
 pending capture after the active production/review work, before delivery completes.
 This does not mark failed captures as complete or waive source-read authorization.
+
+For image-heavy tasks, reuse the image choice in the work-start report. When
+source descriptors already show more than 30 distinct images across the task,
+resolve that question before bulk media acquisition. If counts emerge only from
+capture, stop further image processing to confirm the choice during planning.
+Use explicit image/GIF reference-only policies for exclusions; never substitute
+temporary signed media URLs as permanent public image links.

package/providers/context/resources/procedures/knowledge-review.md

@@ -23,7 +23,7 @@ authorization.
 Without explicit session-managed authority:
 
 - open the report returned by the route;
-- let the user publish or durably omit candidates;
+- let the user approve, reject, or request revisions to candidates;
 - apply the exact copied review code through the returned review apply command; and
 - retain the exact report reference and reviewed scope in this conversation for
   the final completion summary.
@@ -58,10 +58,24 @@ Use the explicit retirement preview after replacement delivery, and repair its
 reported incoming references. Navigation removal alone does not remove content
 from search or packages. A renamed menu does not require a new article identity.
 
-The Review UI names the internal `rejected` decision **Omit** because it is a
-durable content decision, not a request to rewrite the page. When a page needs
-changes, leave that page pending and use `context revise` so the owning
-Author or Composer produces a new Candidate through the same lifecycle.
+The report opens on a change overview, with the existing navigation and an
+expected workspace file tree. The CLI collects approved titles, navigation and
+candidate bodies; it renders their changes mechanically. Do not write a second
+summary in place of candidate content or rewrite the HTML. New workspaces show
+all candidate pages as New. Existing unchanged pages show titles only; changed
+blocks and previous text remain available for comparison. Navigation without a
+Git baseline is labelled as current context, not an invented historical diff.
+
+Approve accepts a candidate. Reject durably omits it; it does not request a
+rewrite or retire an approved article. Entering revision instructions requests
+repair and locks the other choices for that page until cancelled. Bulk approval
+requires confirmation and affects only undecided pages. When the report contains
+new top-level categories, the dialog lists them and requires explicit
+acknowledgment plus an eight-second wait before confirmation is enabled.
+New descendant pages under an existing category do not trigger this extra step. The copied code binds
+the candidate scope, exact content, displayed baseline and revision instructions;
+the CLI validates these together before writing any decision. Follow returned
+repair commands, then obtain review of the repaired candidates.
 
 Use the affected page's `Repair` command in the review material, replacing only
 the correction instruction. Current candidates are repaired within this batch;
@@ -132,3 +146,10 @@ only its own titles. When revising or merging, preserve useful existing detail
 instead of replacing it with generic lookup advice. Repair affected pages or
 use the current planning route for missing topics. These are Agent judgments,
 not minimum article counts, a new coverage ledger or an extra CLI approval gate.
+
+For image-heavy work, show the selected task image policy and actual retained,
+converted and placeholder counts from the candidates and capture reports. Explain
+any fallback affecting reader understanding. Reuse the planning choice; do not
+turn image handling into another per-article approval. Missing image content must
+not be presented as read or fully covered. Build-only media fallbacks are reported
+in the build receipt and do not rewrite approved evidence.

package/providers/context/resources/procedures/knowledge-updates.md

@@ -325,6 +325,12 @@ input without that flag to continue. A preview is not approval and does not make
 a stale revision valid.
 
 Review can approve checked pages while leaving repair pages pending. The HTML
+report shows the current site navigation, candidate changes and expected file
+paths. Unchanged approved pages retain their titles with omitted bodies. Enter
+revision instructions directly in the report; the copied code and following
+instruction lines must be returned together without edits. The CLI binds them to
+the reviewed content and baseline and returns repair commands while leaving
+those pages pending. Changed candidates require a fresh review. The HTML
 review code includes pending positions; managed Review provides the same current
 scope as a JSON template. Send decisions only for pages actually reviewed. Omit
 means a durable exclusion, not repair. Partial approval alone does not build.

package/providers/context/resources/procedures/production-requirements.md

@@ -59,3 +59,26 @@ contains `path`, `question`, `sources` and `batch`, with optional `brief` and
 command to prepare those tasks without a separate investigation-plan submission.
 This does not approve the report or permit writing before user feedback.
 When targets are unclear, use normal preparation and investigate the skeletons.
+
+## Configured coverage and the current request
+
+`src/indexers.yaml` describes the workspace's standing coverage configuration.
+Source registration or presence in that configuration does not, by itself, mean
+that the source needs investigation again for this request. For a bounded document
+addition or revision, identify its reader task and actual evidence dependencies;
+reuse existing approved articles and applicable completed work. Do not broaden a
+document task to unrelated code repositories merely because they are configured.
+
+In planning and completion reports distinguish:
+- work requested and completed in this task;
+- specific content gaps established by reading, with their affected reader tasks;
+- source/environment failures, and whether this task depends on those sources.
+
+A `pending_scopes` count is workflow state, not a count of missing topics,
+repositories to rewrite, or articles to produce. A Git spawn/baseline-read error
+means source availability could not be checked; it does not prove that existing
+knowledge is absent or obsolete. If unrelated configured sources remain blocked,
+say so as a workflow limitation rather than announcing new investigation work.
+Do not delete configured sources, invent exclusions, clear stage files, or mark
+unread material investigated to make the current task appear complete. Follow the
+returned resolution action when the workflow still needs a decision.

package/providers/context/resources/procedures/runtime-event-delivery.md

@@ -22,12 +22,18 @@ remove the Agent host's network approval boundary. Satisfy that boundary by
 requesting host network escalation in the tool invocation that executes the
 flush, not by turning it into another conversational approval gate.
 
-Before requesting network access, run:
+When first requesting host network access, or diagnosing an unknown destination, run:
 
 ```bash
 context logs plan --format json
 ```
 
+When delivery is already authorized and the fixed destination and data policy are
+unchanged, execute `context logs flush --format json` directly. A prior plan
+receipt is not a CLI prerequisite. Do not repeatedly plan or ask for consent on
+retries. `invalid_batch` from the command bridge is local protocol rejection before
+HTTP; repair/update the compatible sink client rather than retrying network access.
+
 The plan is the audit contract. It reports the canonical workspace outbox,
 normally `.tmp/context-runtime/logs/outbox.jsonl`, the event count and kinds,
 the allowlisted property names, the sink command, and the resolved HTTP

package/providers/context/resources/procedures/work-start-report.md

@@ -78,6 +78,42 @@ Estimate timing or savings only when supported by measurements and conditions.
 For a small addition to an existing site, describe only the affected reading
 position; reuse its layout and delivery choices instead of proposing them again.
 
+## Confirm image handling for large tasks
+
+Before bulk image interpretation or article writing, count distinct image resources
+across the current authorized task (all its sources, not each document or batch).
+Use source descriptors and capture reports; do not download images just to count
+or recount unchanged snapshots. If the count exceeds 30 and the user has not
+already selected a policy for this task, include one image-handling question in
+this planning confirmation, in the user's language:
+
+- **Process all (default): intelligent conversion and inclusion.** Read images as
+  needed; retain useful images or convert suitable content into text, tables or
+  diagrams. Conversion is an Agent writing decision supported by actual reading.
+- **Include all:** retain the image presentation; automatic compression and safe
+  processing still apply. An image that cannot be processed within limits becomes
+  a visible placeholder, not an unlimited original-byte delivery promise.
+- **Include none (placeholders):** skip image bytes and interpretation; preserve
+  placeholders and source document links. Do not claim image-content coverage.
+
+Default means the recommended option, not consent after silence. Wait for this
+choice together with the existing plan confirmation; never add per-image or
+per-batch approval. Reuse an explicit choice and any GIF exclusions throughout
+retries. Persist the choice in the existing plan/requirements. If discovery only
+reveals the threshold after initial capture, ask before further image work and
+writing; do not restart completed capture solely to show this question.
+
+Report the known count and unknown resources separately. If the scale was unknown
+and later crosses 30, ask once at that point. At content Review summarize retained,
+converted and placeholder outcomes with reasons, including automatic fallbacks;
+do not ask the same policy question again. A later material policy change belongs
+in the existing plan and normal Review, not a separate approval protocol.
+
+Use `captureLark.resources.images: reference-only` for excluded images and
+`resources.gifs: reference-only` for an explicit GIF exclusion. Both inclusion
+choices use normal capture and package optimization; the Agent controls semantic
+conversion. See the Lark resource guide for configuration and limitations.
+
 ## Review navigation changes in context
 
 Read the knowledge-updates guidance on preserving established navigation intent.
@@ -119,3 +155,39 @@ Reports, plans, skills, candidates, confirmation and transaction process stay in
 `.tmp`. Only formal results and necessary long-term source/requirement decisions
 belong in non-temporary storage. On a new machine, start new production from those
 formal inputs, without requiring the previous report or production session.
+
+## Inputs for the content Review report
+
+Keep approved article titles, paths, reading order and intended placements in the
+normal article plan and knowledge map. Complete their existing structure workflow
+before presenting content Review. The CLI collects those inputs and the actual
+candidate bodies to render the report; do not ask an Agent to fabricate a site
+snapshot, infer approved content from titles, or maintain a second report model.
+For a new workspace, plan the initial categories through the same workflow; the
+report will label its candidate pages New. If it shows unplaced articles, finish
+those placements rather than presenting the fallback as an approved new category.
+Content Review displays changes; it does not replace specific authorization for
+a top-level navigation change already required by this procedure.
+
+## Configured coverage and the current request
+
+`src/indexers.yaml` describes the workspace's standing coverage configuration.
+Source registration or presence in that configuration does not, by itself, mean
+that the source needs investigation again for this request. For a bounded document
+addition or revision, identify its reader task and actual evidence dependencies;
+reuse existing approved articles and applicable completed work. Do not broaden a
+document task to unrelated code repositories merely because they are configured.
+
+In planning and completion reports distinguish:
+- work requested and completed in this task;
+- specific content gaps established by reading, with their affected reader tasks;
+- source/environment failures, and whether this task depends on those sources.
+
+A `pending_scopes` count is workflow state, not a count of missing topics,
+repositories to rewrite, or articles to produce. A Git spawn/baseline-read error
+means source availability could not be checked; it does not prove that existing
+knowledge is absent or obsolete. If unrelated configured sources remain blocked,
+say so as a workflow limitation rather than announcing new investigation work.
+Do not delete configured sources, invent exclusions, clear stage files, or mark
+unread material investigated to make the current task appear complete. Follow the
+returned resolution action when the workflow still needs a decision.