kibi-cli 2.28.1 → 2.30.0

package/src/public/skills/kibi-decisions/resources/decision-records.md

@@ -0,0 +1,142 @@
+# Decision records
+
+Reference for `kibi-decisions`: what the hooks say, the payloads that record
+or change a rule, what Kibi refuses, and the `kb_compile_intent` review loop.
+
+## What the hooks say
+
+| Hook | Text you see | What it means |
+| --- | --- | --- |
+| Prompt submit | `Kibi: this request reads as a rule or decision (it says "<phrase>").` | The phrase matched words such as "must", "never", "always", "from now on", "has to". It only asks whether there is something to record; check with `kb_search`. |
+| Before an edit (Claude Code) | A permission prompt: `Kibi: this edit changes <symbol>, which implements a firm requirement recorded in the knowledge base: <ID>: <title>. Allow it if ...` | The person decides. You are also told: if the edit was declined or could not be confirmed, the requirement stands; if allowed and the change alters what the requirement says, update the requirement with `kb_upsert` in this change. |
+| Before an edit (Codex, Cursor, ZCode) | `Kibi: this edit changes <where>, which implements a firm requirement, a rule the person set: ...` | The host cannot prompt the person, so you must: unless the edit keeps the rule, ask before making it and name the rule. |
+| Stop | `Kibi: this session changed source files and recorded nothing in the knowledge base.` (or the variant naming a rule-like request) | Record the decision the person stated or the change made (a rule, a limit, a default, who may do what) as a requirement linked to its code, or say in one line that there was none. Asked once per session. |
+
+The firm-edit prompt covers current requirements whose effective firmness is
+`firm` and that the edited symbol (or, for a whole-file write, any symbol of
+the file) implements. Superseded and retired requirements never ask. An
+operator can turn the prompt off with `KIBI_CONFIRM_FIRM_EDITS=0`; that does
+not make a firm rule negotiable.
+
+`kb_check` with `sourceFiles` and `includeImpactDiagnostics: true` reports
+`requirement_statement_unverified` (warning) for each current requirement a
+changed behavioral symbol implements, quoting its statement and firmness.
+Kibi never compares code with prose: you do. A change that no longer does what
+the quote says is a conflict with a recorded rule.
+
+## Effective firmness
+
+| `firmness` | `origin.kind` | Effective |
+| --- | --- | --- |
+| `firm`, `default` or `exploratory` | any | as written |
+| absent | `agent` | `default` |
+| absent | `human`, `import`, `migration`, or no origin | `firm` |
+
+`kb_search` answers and `kb_query` rows show the effective value. Hook
+snippets label firm decisions `[firm: ask before changing]` and exploratory
+ones `[exploratory]`; an unlabeled decision is `default`.
+
+## Record a rule the person stated
+
+Discover first (`kb_search`, `kb_query`). If a current requirement already
+says it, link to it instead of restating it. Otherwise write the scenario
+first, then the requirement (the requirement's write sends `specified_by`):
+
+```json
+{"type":"req","id":"REQ-billing-refund-approval","properties":{"title":"Refunds above 500 EUR need a finance approval","status":"open","priority":"must","firmness":"firm","origin":{"kind":"agent","ref":"session request: \"from now on refunds above 500 EUR must be approved by finance\""}},"relationships":[{"type":"specified_by","from":"REQ-billing-refund-approval","to":"SCEN-billing-refund-approval"}],"document":{"body":"Refunds above 500 EUR need a finance approval.\n\n## Context\n\nThe person asked for this rule while changing the refund flow; reason not stated.\n\n## Source\n\n> from now on refunds above 500 EUR must be approved by finance\n\nSession request."}}
+```
+
+Then link the enforcing symbol with `implements` (and its test with
+`covered_by`). Preview with `dryRun: true` when unsure; a passing dry run is
+validation, not the write.
+
+- `origin.ref` is where the words are and the words themselves: a ticket URL,
+  a document path with line, or the session request quoted. Kibi checks that a
+  reference exists, not that it is genuine, so never write words nobody said.
+- `kb_upsert` records `origin.kind: agent` when you omit `origin`; the explicit
+  `firmness: firm` is what makes the rule the person's. Leave
+  `origin.approved_by` to the person.
+- A rule from a document, test or ticket a person wrote is `firm` too, with
+  `origin.ref` to that source.
+- Your own choice is `default`; a guess is `exploratory`. Write the value
+  explicitly whenever you record a decision.
+- For strict modeling (claim keys, predicate facts) follow `kibi-usage`,
+  "Second-agent edits", recipe 1.
+
+## Supersede a rule
+
+Only when the person said to change that rule (firm), or on evidence (default):
+
+1. Write the new scenario, then the new requirement with `supersedes` to the
+   old one (direction new -> old), `firmness: firm` when the old one was firm,
+   `origin.ref` to the person's instruction, and a body section:
+
+   ```markdown
+   ## Why this changed
+
+   The person asked on 2026-10-21 to raise the approval threshold to 1000 EUR
+   because finance now reviews refunds weekly; the old 500 EUR limit predates
+   that review.
+   ```
+
+   At least 12 words, naming the evidence and answering the old decision's
+   reasoning.
+2. Upsert the old requirement with `status: closed` in the same change (its
+   stored `title`, the new status; status and links stay editable on a firm
+   decision).
+3. Re-point or close the old scenario, and move `implements` links to the new
+   requirement when the code now enforces it.
+4. Run `kb_check`.
+
+A preference ("this would be cleaner") is not evidence: propose it in the
+report instead of superseding. An authored requirement is never deleted:
+`kb_delete` returns `outcome: supersession_required`.
+
+## Correct a firm rule in place
+
+When the person asks for a correction that keeps the rule's meaning (a typo, a
+clearer title), update the existing requirement with a new `origin.ref` citing
+their words. Without new human words, Kibi refuses to change a firm decision's
+title, statement (or an ADR's body) or to lower its firmness.
+
+## What Kibi refuses or blocks
+
+| Where | Code or message | Fix |
+| --- | --- | --- |
+| `kb_upsert`, `dryRun`, `kb_apply_plan` step | `Entity validation failed: <ID> is a firm decision, so kb_upsert does not change its <fields> in place.` | Ask the person; then cite their words in `origin.ref`, or supersede with a firm decision. |
+| `kb_upsert` | `firmness: firm` without `origin.ref` (or `approval_ref` on a requirement) | Add the reference to the person's words. |
+| `kb_upsert` | `firmness is only valid on req and adr entities` | Remove `firmness` from facts, scenarios, tests and symbols. |
+| `kb_check` | `decision-firm-unreferenced` | As above. |
+| `kb_check` | `decision-firm-superseded` | The superseding decision must be `firm` too. |
+| `kb_check` | `decision-firm-edited` | A firm decision's claim changed in the working tree since `HEAD` without new words; revert, or cite their words. |
+| `kb_check` | `supersedes-reason-missing` | Add `## Why this changed` (12+ words). |
+| `kb_check` | `superseded-requirement-open` | Close the superseded requirement. |
+
+## The kb_compile_intent review loop
+
+`kb_compile_intent` (`mode: "create"` or `"update"`, the complete intent,
+`context` when creating) is read-only and returns a plan.
+
+- `contradictionAnalysis.outcome`:
+  - `no_conflict`: the check could decide and found none.
+  - `conflict`: a witness or your own verdict says the intent contradicts a
+    current requirement. The plan `status` is `blocked`.
+  - `unresolved`: Kibi could not decide. `reasons` say why
+    (`unmodeled_target`, `unmodeled_related`, `analysis_unavailable`). Never
+    read it as consistent.
+- `reviewCandidates[]`: up to five current requirements Kibi could not compare
+  logically, each with `requirementId`, `title`, `sentence` (the closest
+  sentence), `statement`, `score`, `modeled` and `review` (`pending`,
+  `conflict` or `no_conflict`).
+- Compare the intent with each candidate's `statement`, then compile again
+  with the same input plus
+  `conflictReviews: [{"requirementId": "...", "verdict": "conflict" | "no_conflict", "note": "..."}]`.
+  A verdict may also name a requirement discovery did not rank.
+- On `conflict`: ask the person which rule holds. Then supersede or change
+  the requirement (as above), or change the intent, and compile again.
+- Plan `status` is `ready`, `needs_resolution` (ambiguous or unmodeled
+  propositions, missing targets) or `blocked`. Apply only a `ready` plan the
+  person approved, with `kb_apply_plan` and its full `planHash`.
+- A compiled plan records the new requirement as an agent's. When the intent
+  is the person's stated rule, follow the apply with a `kb_upsert` that sets
+  `firmness: firm` and `origin.ref` on that requirement.

package/src/public/skills/kibi-decisions/resources/signoff-views.md

@@ -0,0 +1,66 @@
+# Signoff views
+
+`kibi signoff` renders what a person signs off, so the agent never retypes
+requirement text from memory. It is a CLI command, not an MCP tool.
+
+| View | Command | When |
+| --- | --- | --- |
+| Decisions to confirm | `kibi signoff --questions <file or ->` | Mid-work: rules to record as firm, a change to a recorded rule, a conflict |
+| Requirements signoff | `kibi signoff` (`--staged` for the index) | End of work: before asking to commit |
+
+Both print plain Markdown by default (`--format chat`): headings, quoted
+statements and lists that render in any chat or messaging surface. Put it in
+your reply unchanged: never wrap it in a code block, summarize it or retell a
+requirement in your own words. Every run also writes
+`.kb/briefs/signoff/latest.json`, `latest.md` and `latest.html` (git-ignored);
+`latest.html` is a self-contained page with the full view.
+
+## Showing it on your host
+
+| Host | Best view |
+| --- | --- |
+| Claude Code, Codex, Cursor, ZCode, VS Code chat | Paste the default output; `/kibi-signoff` (Claude Code, Cursor, ZCode) runs it. Offer `kibi signoff --format html --open` for the page |
+| OpenCode terminal | `kibi signoff --format md` and paste the fenced block unchanged: kibi-opencode draws it as a card, and `/kibi-signoff` reopens it |
+| Replies sent through a messaging tool (cloud or project threads, chat bots) | Pass the default output as the message text, and attach or publish `.kb/briefs/signoff/latest.html` next to it when the tool can carry a file or a page |
+| VS Code or Cursor with the kibi-vscode extension | The **Kibi: Show Requirements Signoff** panel follows every run |
+| Remote machine, no browser | Give the person the path of `latest.html` |
+
+`--format md` wraps the same Markdown in a ```` ```kibi-signoff ```` or
+```` ```kibi-questions ```` fence for host renderers; use it only where one
+draws it, because everywhere else a fence shows as raw code.
+`KIBI_SIGNOFF_FORMAT` sets the default for a host. `--format html --open`
+writes and opens the page; `--format json` prints the view.
+
+## Questions input
+
+`{"questions":[...]}` or a bare array. Each question:
+
+| Field | Required | Meaning |
+| --- | --- | --- |
+| `statement` | yes | The rule as it would be recorded |
+| `title` | no | A short name for the card |
+| `id` | no | The recorded requirement the question would change; the view quotes its current statement and firmness |
+| `quote` | no | The person's words that would go in `origin.ref` |
+| `firmness` | no | `firm`, `default` or `exploratory`, as it would be recorded |
+| `governs` | no | Files or symbols the rule governs |
+| `conflictsWith` | no | Recorded requirement ids it conflicts with; the view quotes them |
+| `question` | no | What you ask, when it is more than "record this?" |
+| `answer` | no | The person's answer; answered questions are struck out |
+
+The person answers by number ("1 yes, 2 exploratory, 3 reword: ..."). Copy
+each answer into `answer`, render again, and record the rules only once every
+question is answered. An answer is the person's words: quote it in
+`origin.ref`; a view itself is never an approval.
+
+## Summary contents
+
+| Section | Meaning |
+| --- | --- |
+| New | Requirement documents the change adds |
+| Changed | Rewritten requirements; a changed statement shows before and after |
+| Affected by the code change | Unchanged requirements whose implementing code changed (the `requirement_statement_unverified` set) |
+| Superseded or retired | Requirements the change retires, supersedes or deletes |
+
+Each requirement shows its firmness, origin, superseded rules, changed code
+symbols, scenarios and proof state (the earliest unmet proof gate). Parts
+Kibi could not compute are named in a `Note:` line instead of failing.

package/src/public/skills/kibi-freshness/SKILL.md

@@ -53,6 +53,8 @@ echo '{}' | npx --no-install kibi status --input -
   `committed_with_repairs`, follow required typed `nextActions`; never retry the
   original operation.
 
+For every freshness, branch, or completion question, call `kb_status` first, then `kb_query` the entities and source files that status names, then `kb_check` as the last Kibi call before answering. Never skip the query step, even when status alone seems to answer the question. If another query or write follows a check, run `kb_check` again before answering. When asked whether requirements are consistent or proven, run `kb_query`, then `kb_check` with the domain-contradictions rule, then `kb_coverage`, and read whether the analysis was complete. Report any unmodeled or unknown clause as incomplete analysis, not as consistent. Only when every clause is fully modeled and compatible, state plainly that no conflict was found. For authorized knowledge changes, keep the exact supplied target and intended relationship edges, and discover and query those targets first. Validate the payload with `kb_upsert` and `dryRun: true`, then perform the authorized `kb_upsert` with that same payload and read the result back exactly before the final checks. A passing dry run is validation, not completion. Distinct authorized payloads may need several sequential writes, each validated and read back, but never retry a mutation that already committed. If a mutation returns `committed_with_repairs`, follow its typed `nextActions` instead of repeating the original operation.
+
 ## Guidance
 
 - Treat MCP and the trusted project-local CLI as equal peer surfaces.

package/src/public/skills/kibi-traceability/SKILL.md

@@ -29,6 +29,8 @@ interfaces. Select the surface by capability; neither is semantically preferred.
 For CLI JSON, use a dedicated route such as
 `printf '%s\n' '{}' | kibi status --input -`.
 
+For every traceability answer, begin with `kb_search` even when an identifier is already named, then exact-filter with `kb_query`, inspect relationships with `kb_graph`, and make `kb_status` the final Kibi call, including after `kb_graph` or `kb_coverage`; report no conclusions until that closing `kb_status` returns. When judging scenario feasibility, ground each step in the queried requirements, scenarios, tests, and symbols, and state missing links, ambiguity, or stale evidence explicitly rather than inferring them. For authorized coverage or link work, keep the exact supplied target and intended edges, discover and query those targets, then validate the payload with `kb_upsert` and `dryRun: true`. A passing dry run is only validation, not completion: next perform the real `kb_upsert` with the same payload, exact-read the result back with `kb_query`, run `kb_check` and `kb_coverage`, report the resulting proof state, and finish with `kb_status`. Distinct authorized payloads may each need their own sequential dry run and write, so there is no single-write limit, but never retry a mutation that already committed; on `committed_with_repairs`, execute the typed repair action instead.
+
 ## Workflow
 
 Use the visible approved MCP tools or trusted project-local CLI as equal peer
@@ -50,6 +52,8 @@ plan_returned` with a hash-bound plan whose `nextActions[0].input` is the exact
 through a new requirement linked with `supersedes`. Relationship deletes patch
 shards while preserving unrelated records.
 
+For executable coverage work, keep the production symbol's implements link to its requirement and link that same production symbol to its qualifying test with covered_by; reserve executable_for for a symbol that stands for the executable test itself, and never put executable_for on a symbol that also carries production ownership or coverage. Preserve the exact supplied target and intended edges, discover and query those targets, validate the payload with `kb_upsert` and `dryRun: true`, then perform the authorized real `kb_upsert` with the same payload and exact-read the result with `kb_query` before final checks, because validation alone is not completion. A passing dry run does not guarantee the commit: if the real `kb_upsert` returns an error, read its message and correct only the conflicting relationship on the same target rather than working around it with a new entity, then validate again with `dryRun: true`, write again, and exact-read it back. Distinct authorized payloads may each need their own sequential validation and write, so there is no single write limit, but never retry a mutation that already committed; on `committed_with_repairs`, execute the typed repair action instead. Then run `kb_check` and `kb_coverage`, report the proof and verification state they return, treating a dirty verification snapshot after a write as expected rather than as a failure, and finish with `kb_status`.
+
 ## Typed recovery and proof
 
 Parse `kibiProtocol`, `resultVersion`, `status`, `effects`, diagnostics, and

package/src/public/skills/kibi-usage/SKILL.md

@@ -2,7 +2,7 @@
 id: kibi-usage
 name: Kibi Usage
 description: Use Kibi's source-first, exact-Git, migration-aware, proof-aware operations safely across MCP or the trusted local CLI, including partial completion repair.
-version: 2.20.0
+version: 2.22.0
 kibiCompatibility: ">=1.0.0"
 tags:
   - kibi
@@ -121,6 +121,10 @@ component that renders it) with `implements` through `kb_upsert`. Without a read
 still answers and `kibi sync` regenerates the manifest. The search tip and the
 `.kb/` note appear once per session and need no action.
 
+Before you edit a source file that Kibi tracks, run `kb_search` with that file in `sourceLocations`, and treat both its ranked results and its `data.answer` layer as discovery only, not as the governing text. Then call `kb_query` by `id` on each current governing requirement that the search names, read what each one requires, and make the edit only after those exact reads. A `kb_check`, `kb_status` or `kb_coverage` result does not replace that `kb_query` read. This rule only adds an exact read before the edit and does not change the order of any other step in this workflow. It adds nothing to a modeling or write task: that task still ends with the real `kb_upsert` of every planned fact and requirement and the closing `kb_check`.
+
+For an edit that keeps the governing rule, limit the consultation before the edit to exactly two Kibi calls: `kb_search` with the file in `sourceLocations`, then one `kb_query` (by the governing requirement's `id`, or by `sourceFile` when several requirements govern the file). Do not start with `kb_status`, and do not repeat the search or the query before editing; run `kb_status`, `kb_check` and `kb_coverage` after the edit. This limit applies only to source edits. A task that models a claim keeps its full order, one step after the other: `kb_search`, then the exact `kb_query` of the supplied target, then `kb_model` with `mode: "analyze"` on the complete claim, then `mode: "predicates"` for each relational clause, then `kb_model` with `mode: "requirement"` for a strict or typed-rule clause, and only then the real `kb_upsert` of every planned fact and requirement and the closing `kb_check`. Never call `kb_model` before that exact `kb_query`, and never stop at a dry run.
+
 ## Safe workflow
 
 1. Always discover before you mutate: start with `kb_search`, then exact-filter with `kb_query`. Use
@@ -356,7 +360,9 @@ authored. Each one is the order the write path accepts.
 
 Entities record `origin` (`kind`: `human`, `agent`, `migration`, or `import`,
 plus optional `ref` and `approved_by`). `kb_upsert` records `kind: agent` on
-entities you create; leave `origin.approved_by` to the person who reviews them.
+entities you create; leave `origin.approved_by` to the person who reviews them,
+and ask for that review with `kibi signoff` (see `kibi-decisions`), never with
+a summary in your own words.
 An exception (a requirement that `exempts` another) exempts nothing until a
 human sets `approved_by`; `kb_check` flags approvals an agent recorded without
 corroboration. Never fill in an approval yourself. Record why a requirement
@@ -365,6 +371,8 @@ exists in `rationale`, or link the ADR that explains it.
 ## Decision firmness
 
 Requirements and ADRs carry `firmness`, which says what you may do with them.
+Load `kibi-decisions` when a person states a rule, a hook asks before an edit
+of firm code, or a request conflicts with a recorded rule.
 Search answers, `kb_query` rows and hook snippets show it (snippets label only
 `[firm: ask before changing]` and `[exploratory]`; an unlabeled decision is
 `default`).
@@ -394,17 +402,12 @@ task or in reply: then supersede it with `firmness: firm` and `origin.ref` to
 their words (or, for a correction they asked for in place, update it with
 `origin.ref` to those words), and say in the final answer which rule changed.
 
-Set `firmness` from what a human actually said: only their stated rule is
-`firm` (with `origin.ref`); "do whatever" or your own pick is `default`; a
-guess is `exploratory`. Write it explicitly whenever you record a decision,
-including when importing rules from documents, tests or tickets (a rule a
-person wrote there is `firm`, with `origin.ref` to that document). A
-supersede needs `## Why this changed` in the new entity's body (at least 12
-words): name the evidence (a failing test, a measured cost, a constraint the
-old decision missed, a new requirement, the person's instruction) and answer
-the old decision's reasoning; `kb_check` blocks it otherwise. A preference is
-not evidence: propose it instead of superseding. Kibi checks only that a firm
-decision has a reference; never cite words nobody said.
+Set `firmness` explicitly from what a human actually said: only their stated
+rule is `firm` (with `origin.ref` to their words); your own pick is `default`;
+a guess is `exploratory`. A supersede needs a `## Why this changed` section of
+at least 12 words naming the evidence; a preference is not evidence. Never
+cite words nobody said. `kibi-decisions` has the recording and supersede
+payloads.
 
 Conditional clauses ("only when", "only if", "must not ... unless") compile
 into typed `forbid` rules linked by `requires_rule`, not property facts; one
@@ -418,7 +421,8 @@ read `unresolved` as consistent. Compare the intent with each candidate's
 `statement`, then compile again with `conflictReviews`
 (`{requirementId, verdict: "conflict" | "no_conflict", note}`). On a conflict
 the plan is `blocked`: ask the human which rule holds (a firm requirement is
-theirs), then supersede or change it, or change the intent.
+theirs), then supersede or change it, or change the intent (see
+`kibi-decisions`).
 
 ## Semantic and proof guardrails
 
@@ -558,4 +562,5 @@ retrying the same payload; inspect the typed error and `nextActions`.
 Kibi may now author tracked Markdown/YAML evidence transactionally. The former
 statement “Kibi operation writes do not automatically stage markdown evidence”
 means Kibi writes do not Git-stage it: review and commit remain ordinary Git
-workflows.
+workflows. When `kibi check --staged` or the pre-commit hook blocks a commit,
+load `kibi-commit-review`.