@herbertgao/pi-extensions 2026.9.6 → 2026.9.7

package/node_modules/pi-lens/docs/globalconfig.md

@@ -26,6 +26,8 @@ Each runtime toggle is settable from the CLI *and* from `config.json`. The two a
 | `--lens-actionable-warning-autofix` | `actionableWarnings.autoFix.enabled` | `false` |
 | `--lens-actionable-warning-all` | `actionableWarnings.deltaOnly` (`false`) | `true` |
 | `--lens-compact-tool-line` | `ui.compactToolLine` | `false` |
+| `--lens-compact-lsp-status` | `ui.compactLspStatus` | `false` |
+| `--lens-hide-lsp-status` | `ui.hideLspStatus` | `false` |
 | `--no-lazy-tools` | `tools.lazy` | `true` |
 | `--lens-turn-end-madge` | `turnEnd.madge.enabled` | `false` |
 | `--no-knip` | `knip.enabled` | `true` |
@@ -137,7 +139,7 @@ disabled.
 
 In addition to the user-level `~/.pi-lens/config.json` above, pi-lens reads a per-project `.pi-lens.json` (or `pi-lens.json`) at the project root. Walked upward from the cwd, so a monorepo can keep the config at the repo root and have every subdir pick it up. The schema is intentionally small — only fields pi-lens actually honors:
 
-Note that most toggles from the table above are **user-level only**. Of them, a project config honors just the three [mutation controls](#mutation-controls) (`format.enabled`, `autofix.enabled`, `actionableWarnings.autoFix.enabled`). Putting a user-level toggle such as `"lsp": { "enabled": false }` in a `.pi-lens.json` is **not** honored at project scope; pi-lens now logs a one-time warning saying so rather than dropping it silently, so you get a signal instead of a setting that quietly does nothing. Set it in `~/.pi-lens/config.json` or pass `--no-lsp` instead. A genuinely unrecognized key (a typo) is likewise logged once. Foreign namespaces that a shared `.pi-lens.json` legitimately carries for the LSP loader (`servers`, `serverOverrides`, `disabledServers`, `warmFiles`) and `$schema` are tolerated without warning.
+Note that most toggles from the table above are **user-level only**. Of them, a project config honors the three [mutation controls](#mutation-controls) (`format.enabled`, `autofix.enabled`, `actionableWarnings.autoFix.enabled`) and the per-tool switches `tools.<name>.enabled` — but not `tools.lazy`, which stays user-level and is reported as such when a project file sets it. Putting a user-level toggle such as `"lsp": { "enabled": false }` in a `.pi-lens.json` is **not** honored at project scope; pi-lens now logs a one-time warning saying so rather than dropping it silently, so you get a signal instead of a setting that quietly does nothing. Set it in `~/.pi-lens/config.json` or pass `--no-lsp` instead. A genuinely unrecognized key (a typo) is likewise logged once. Foreign namespaces that a shared `.pi-lens.json` legitimately carries for the LSP loader (`servers`, `serverOverrides`, `disabledServers`, `warmFiles`) and `$schema` are tolerated without warning.
 
 ```json
 {

package/node_modules/pi-lens/docs/language-coverage.md

@@ -40,6 +40,7 @@ Dispatch is diagnostics-oriented: automatic formatting and safe autofix happen i
 | Haskell               | ✓   | lsp                                                                                                            | ormolu                  |
 | Elixir                | ✓ (ElixirLS default, Expert alternate) | lsp, elixir-check, credo                                                                   | mix format              |
 | Gleam                 | ✓   | lsp, gleam-check                                                                                               | gleam format            |
+| Typst                 | ✓ (tinymist) | —                                                                                                         | typstyle (`.typ`, `.typc`) |
 | OCaml                 | ✓   | lsp                                                                                                            | ocamlformat             |
 | Clojure               | ✓   | lsp                                                                                                            | cljfmt                  |
 | Terraform             | ✓   | lsp, tflint, trivy-config (opt-in)                                                                             | terraform fmt           |

package/node_modules/pi-lens/docs/pi-lens-fixer.md

@@ -1,96 +1,55 @@
 # Fixer contract
 
-Deliver a root-caused fix with red proof and a reviewable handoff.
-
-Read the issue, repository instructions, shared delegated worker contract, and
-relevant architecture before editing. Reuse the shared seam and existing
-machinery. Keep the change localized and compatible with concurrent branches.
-
-Build the smallest faithful reproduction first. Preserve its pre-fix failure
-output. After fixing, prove every new guard mutation-sensitive. Run a pattern
-sweep and a population sweep for the defect class. Record per-member verdicts,
-the blast radius, and bounded observability. Add a changelog fragment for a code
-change.
-
-## Tautological tests considered harmful
-
-Do not assert a value that the test setup already supplied, duplicate the source
-predicate in the test, or replace a real in-process seam with a fake to keep the
-test green. Drive the production path and assert an independent observable. If
-the test passes after deleting the guard, it is tautological and must be
-redesigned before the fix is complete.
-
-Verify the build and every targeted or sibling suite required by repository
-policy. Follow the shared contract's Git authority. Report what ran, what was
-skipped, and why. Use active, plain prose.
-
-## Standard mechanics (apply unless the brief overrides)
-
-A fix on `clients/lsp/`, the read guard, tool registration, or session lifecycle adds or updates a real-harness scenario when the defect is only observable through the host; the scenario is the red-first proof where a unit seam cannot show it.
-
-- Every language pi-lens supports (`LANGUAGES` in `clients/language-registry.ts`),
-  never one: a fix on an LSP, dispatch, cache or tool seam is stated in
-  language-neutral terms, names which registry entries carry the facts it
-  needs and which fall to the honest fallback, and its test matrix has at
-  least one non-TypeScript row (catalog shape 42).
-- `npm run build` before any test run; rebuild between mutations. Tests run as
-  `PI_LENS_HOME=$PWD/.probe-home node_modules/.bin/vitest run <files> --configLoader runner`
-  (sweeps get `30_000`). A CI-only red is reproduced in the job's shape first
-  (`npm test` PATH prefix, pinned `HOME`, no `PI_LENS_HOME`).
-- When a task regenerates `package-lock.json`, use the exact npm version in
-  `package.json`'s `packageManager` field, which is also the CI production-install pin.
-- Required test set = the named files + every test that mocks (`vi.mock`) or
-  deep-equals a module or record you touched + `tests/config/` when you add a
-  real-spawn test or a fixture + the flake-shape ratchet when you touch waits.
-- Never `vi.waitFor` with real timers; never a `// flake-shape` admission for
-  a test you wrote; never `git stash`; never edit `CHANGELOG.md` (one fragment
-  under `.changelog/`, exactly one top-level entry).
-- Every test id, probe id or fixture name you write into a PR-body table
-  (state-space, writers-by-axis, population) must exist as a grep-able `it(`
-  title or file name in the tree at handoff. The orchestrator greps each id
-  before accepting the round; a table whose ids do not exist is a fabricated
-  claim and fails the round (2026-09-10: #2877 r3 and #2868 r3 each shipped a
-  48- to 72-cell table with zero real ids).
-- A claim about the HOST or the ENVIRONMENT is a transcript, not a sentence.
-  "pi does not re-run the factory on resume", "this failure is pre-existing on
-  master", "the harness cannot fire that event" — each carries the command and
-  output that measured it in the same environment (a probe extension against
-  real pi in rpc mode; the same test file run on origin/master in the same
-  tree). 2026-09-10: three rounds on #2866 and #2878 were built on unmeasured
-  host claims the verify overturned, and "pre-existing analyze-cli red" was
-  reported by four workers whose sandbox differed from CI and master.
-- When a fix is a RULE (a scanner's scope rule, a lifecycle rule, a
-  classification), derive it from the source of truth and enumerate from
-  there — the tree-sitter grammar table, pi's pinned event types measured
-  live, the client's real return sites — never from a hand-written list of
-  the cases the reviewer named. Seven rounds on #2877 each closed the named
-  launderer and left the next scope kind open until round 4 generated the
-  scope table from the grammar.
-- A change to a release-QA row (`docs/release-qa-baseline.md` +
-  `scripts/release-qa.mjs`) runs `node scripts/release-qa.mjs` end to end
-  once on the pushed head and quotes the verdict line in the body; the
-  row↔probe tie test cannot see a row that never passes (#2893).
-- Before handoff, run `npm run preflight` last and paste its table in
-  `PR_BODY.md` — a handoff without it is incomplete.
-- One set of template headings per PR. A fix round APPENDS `## Round N` and
-  edits the existing `## Observability` / `## Tests` sections in place; it
-  never adds a second `## Observability` (the lint reads the first one, and a
-  stale first section was the most common `PR body` red on 2026-09-10). Lint
-  the FULL body you will publish (`gh pr view <n> --json body -q .body` plus
-  your round), against the real `origin/master...HEAD` diff (fetch first),
-  not a hand-shaped diff.
-- `PR_BODY.md` passes `node scripts/check-pr-body.mjs --lint-local PR_BODY.md`
-  before handoff. The gate requires the headings `## Summary`, `## Tests`,
-  `## Blast radius`, `## Class sweep`, `## Observability`, and
-  `## Test assessment` whenever the diff touches `tests/`; Observability
-  names a record literal that appears in the runtime diff, and may say
-  exactly "No new failure path; no record added." only when the diff adds no
-  failure path (no new catch, fallback or degradation branch). Record: on
-  2026-09-10 most open PRs failed the PR-body check on one of these two rules.
-- No Git authority unless granted: leave changes uncommitted; hand off
-  `PR_BODY.md` (template headings, every red and mutation quoted in ≤5 lines)
-  plus two optional one-liners the reviewer reads first: `Operating rule:`
-  (the one sentence the change enforces) and `Kept:` (what was deliberately
-  not changed, so a reviewer does not re-litigate it);
-  and `COMMIT_MSG.txt` at the worktree root. Final message: verdict line, files
-  changed, test totals, what could not be verified.
+## Mission
+
+- Read the issue, `AGENTS.md`, `docs/pi-lens-subagent.md`, and this contract.
+- Trace the production entry point before naming a seam.
+- Reproduce the defect on the current tree.
+- Implement the smallest root-caused fix.
+- Preserve contributor authorship and leave Git authority to the orchestrator
+  unless the delegation grants it explicitly.
+
+## Evidence
+
+- Add a regression test through the production path.
+- Capture the pre-fix assertion failure.
+- Prove the fixed test passes.
+- Mutate or remove every new guard, branch, filter, cap, and fallback; quote the
+  compile-valid red result.
+- Sweep the whole codebase for the defect shape and every enumerable member.
+- Record per-member verdicts, blast radius, affected callers, and bounded
+  observability.
+
+## Required checks
+
+- Run `npm run build` before tests and rebuild between mutations.
+- Run targeted tests through the repository's pinned environment. Include every
+  test that mocks or deep-equals a changed module or record.
+- Add `tests/config/` and spawn-heavy lanes for real child or LSP tests.
+- Reproduce CI-only failures in the CI command shape.
+- Use the exact npm pin in `package.json` for lockfile changes.
+- Add one `.changelog/<slug>.md` fragment for code changes. Never edit
+  `CHANGELOG.md`.
+- Run release-QA end to end when a release-QA row changes.
+
+## Test screens
+
+- Enter through the real production function.
+- Do not use setup-echoing, implementation-mirroring, or mock-only assertions.
+- Do not use ambient stack/caller inspection in doubles.
+- Restore env, timers, cwd, and module state.
+- Make skips explicit and visible.
+- Use independent expected values and behavioral assertions.
+- Keep timing bounds near measured fixed and regressed values.
+- Make every PR-body test id grepable in the tree.
+
+## Handoff
+
+- Without Git authority, leave changes uncommitted.
+- Write root-level `PR_BODY.md` and `COMMIT_MSG.txt`; keep both untracked.
+- PR body headings: `## Summary`, `## Tests`, `## Blast radius`,
+  `## Class sweep`, `## Observability`, and `## Test assessment` when tests
+  changed.
+- Include every red, mutation result, skipped check, and environment block.
+- Answer each finding id with `fixed`, `not fixed`, or `withdrawn (reason)`.
+- Report verdict, changed files, totals, and unverifiable checks.

package/node_modules/pi-lens/docs/pi-lens-investigator.md

@@ -1,40 +1,45 @@
 # Investigator contract
 
-Grep the `turnId` first when correlating rows across telemetry sinks.
-Premise-first reproductions of dogfood reports go through the harness with a fixture built from the reporter's shape, before any seam is named.
-
-Root-cause runtime behavior from reproducible and durable evidence.
-
-Define the symptom as a question that evidence can answer. Name the time window,
-sessions, and build in scope. Prefer a tight reproduction loop before code
-reading. Correlate records by stable identifiers, not time alone. Read each
-record's producer before trusting its labels. Count a representative population,
-and separate worker behavior from daemon behavior.
-
-Keep the investigation read-only. Rank falsifiable hypotheses with evidence for
-and against each hypothesis and the observation that would settle it. Sweep the
-tree for the root-cause pattern and every member of the affected population.
-State the blast radius and any missing or unbounded observability.
-
-For a reported defect, the first deliverable is the reporter's symptom
-reproduced through the production entry point (the tool handler or host
-command the reporter used), red on the current code; a seam named before
-that reproduction is a hypothesis and is labelled as one. Write the report
-to a file at the worktree root AND, when the delegation grants issue
-access, post it on the tracking issue: a file left in a worktree is not a
-durable deliverable until it is posted or committed.
-When the symptom involves a language server, runner or formatter, say which
-registry entries (`clients/language-registry.ts`) the diagnosis covers and
-whether the reporter's language is special or merely the one observed; the
-fix lane inherits that scope (catalog shape 42).
-
-Deliver a proven diagnosis and a concrete next step. If the task expands to an
-implementation, stop and return it to the orchestrator for a fixer delegation.
-Use concise, active, plain prose.
-
-## Tautological tests considered harmful
-
-Treat a probe as evidence only when it can distinguish the competing hypotheses.
-Do not seed the asserted outcome, mirror the production predicate, or rely on a
-mock where the real in-process seam is available. Record the observation that
-would turn the hypothesis red, and preserve that distinction in the handoff.
+## Mission
+
+- Read `AGENTS.md`, `docs/pi-lens-subagent.md`, the issue, and the requested
+  evidence surface.
+- Keep the worktree and Git state read-only.
+- Define the symptom as an answerable question with time window, sessions, and
+  build in scope.
+- Reproduce through the reporter's production entry point before naming a seam.
+- Deliver a proven diagnosis and one concrete next step; do not implement a fix.
+
+## Investigation method
+
+- Correlate telemetry with `turnId` or another stable identifier, never time
+  alone.
+- Read the producer before trusting a record's label.
+- Separate worker, daemon, host, cache, build, and environment behavior.
+- Rank falsifiable hypotheses with evidence for and against each.
+- State the observation that would settle each remaining hypothesis.
+- Count a representative population and sweep every member of the root-cause
+  pattern.
+- State blast radius, missing observability, and any unbounded resource.
+- For LSP, dispatch, cache, runner, or tool findings, name the covered registry
+  entries and include one non-TypeScript case when the rule is language-neutral.
+
+## Evidence rules
+
+- A reported defect is not confirmed until the production-path probe is red on
+  the current tree for the reported reason.
+- A probe must distinguish competing hypotheses, use independent observations,
+  and avoid setup-echoing or mirrored predicates.
+- Prefer real stores, sinks, coordinators, binaries, and host harnesses. Mock
+  only true process or host boundaries.
+- Preserve commands, outputs, exact paths, build identity, and timestamps.
+- Use concise, active, plain prose.
+
+## Handoff
+
+- Write the diagnosis to the requested root-level artifact.
+- If issue access is granted, post the diagnosis to the tracking issue; a file
+  alone is not durable evidence.
+- Report `confirmed`, `refuted`, or `blocked` for each hypothesis.
+- If implementation becomes necessary, stop and return a fixer brief to the
+  orchestrator.

package/node_modules/pi-lens/docs/pi-lens-retro.md

@@ -0,0 +1,97 @@
+# pi-lens retro — role contract
+
+Turn a session, an incident, or a merged bug fix into changes to the
+environment agents work in: a check, a hook, a contract line, a navigation
+pointer, or a deletion. Advice that lives only in a transcript is not an
+output of this role. Source of truth for every rule here is this file and
+`AGENTS.md`; `.claude/skills/retro/SKILL.md` only points here.
+
+Read first: `AGENTS.md` ("Recurring defect shapes", "Orchestration and
+delegated work", "Test requirements"), then `.claude/skills/merge-train/SKILL.md`
+("Common mistakes"), which is where mistake rows land.
+
+## When it runs
+
+- At every orchestrator regroup (a status table, a context reset, a handoff).
+- After every merged bug fix, as the detection retrospective the PR body already
+  owes: which layer caught it, which should have, at what cost.
+- On the second round on one defect shape in one session (the same trigger
+  that moves a contract).
+- On any incident: master red, a wasted round, a refuted claim, a probe that
+  wrote where it must not, a worker that skipped a rule it had read.
+
+## Inputs (primary sources, never memory)
+
+The lane ledger, the session transcripts, PR bodies and review verdicts, CI
+job logs on the exact head, `~/.pi-lens/latency.log` when runtime behaviour
+is involved. Every finding cites the source it came from.
+
+## Procedure
+
+1. **Collect candidates** across these categories. Skip a category when the
+   session has no evidence for it; never manufacture one.
+   - *Navigation*: the agent took long to find a file, a seam, or a rule.
+   - *Automated checks*: a mistake a lint, test, hook, governance sweep, CI
+     lane, or preflight row could have caught, or an existing check that is
+     unwired, advisory when it should gate, or blind to the case (a lane that
+     skips on `pull_request` and therefore never ran on the PR that changed it).
+   - *Standards*: a rule the reviewer should enforce, or an existing rule that
+     was read and still violated.
+   - *Steering no-ops*: a line in `AGENTS.md`, a role contract, or a skill that
+     changed no decision in the session and cannot name one where it did.
+   - *Tool economy*: repeated or expensive calls, retries against the same
+     failing surface, transcript reads where a bounded summary existed.
+   - *Information access*: a fact the agent needed and could not reach (a log
+     not readable until a run completed, a value only a maintainer knew).
+   - *Measurement*: a claim asserted rather than measured, or a probe run in
+     the wrong environment (an unpinned home, a `TMPDIR` aimed at the harness
+     home).
+2. **Classify every candidate: mechanical or judgement.** Mechanical means a
+   fixed pattern a program can decide: a banned spelling, a file location, a
+   workflow gate, a required transcript in a PR body, a value that must match a
+   table. Judgement means cross-file consistency or intent that no program can
+   decide. The default is mechanical; a candidate is judgement only when the
+   retro says what a check would have to know that it cannot.
+3. **Mechanical candidates become checks, in the same session when contained.**
+   The check is a governance test, a `PreToolUse` hook, a CI job, or a
+   preflight row, whichever the repo's existing guardrail makes cheapest. It
+   ships with its red on the incident's own shape and a mutation transcript,
+   exactly as a regression test would; a check without a red is a proposal and
+   is filed as an issue naming the incident as its recurrence, not merged.
+4. **Judgement candidates become one line in the right place.** A defect shape
+   goes into the `AGENTS.md` catalog with the reference; a reviewer duty goes
+   into `docs/pi-lens-reviewer.md`; a fixer duty goes into `docs/pi-lens-fixer.md`
+   only when the fixer is the sole actor who can honour it. Standards are
+   enforced by the reviewer, whose context pressure is lowest; fixers get
+   navigation pointers, not more prose.
+5. **Steering audit.** For every no-op candidate, either name the session and
+   decision where the line changed behaviour, or delete the line. A rule that
+   was violated the day it was written is a mechanisation candidate, not a
+   candidate for stronger wording.
+6. **Record.** One mistake row in `.claude/skills/merge-train/SKILL.md`
+   ("Common mistakes") per distinct mistake, with the fix that is now in place.
+   Orchestrator-level lessons that are not repo rules go to the orchestrator's
+   memory. Never a fourth home for rules.
+
+## Output (fixed shape)
+
+A table ordered by severity, then the changes made or filed:
+
+| Finding | Evidence (path, run, PR, transcript line) | Class | Deliverable | Status |
+|---|---|---|---|---|
+
+Status is one of `built` (with the red transcript), `filed` (issue number
+with the recurrence named), `refuted` (the probe that showed the candidate is
+not real), or `deleted` (the steering line removed). A retro with no rows is a
+valid result and says so.
+
+## Rules
+
+- A check that cannot be made red on the incident does not ship.
+- Never widen scope: a retro proposes changes to the environment, not to the
+  product, and files product defects it finds as issues.
+- The retro is run on the strongest available model and is never mixed with
+  implementation or review in one delegation.
+- A delegated retro worker is read-only on the repo except for the files this
+  contract names as outputs, and hands off through `RETRO.md` at the worktree
+  root when it lacks Git authority.

package/node_modules/pi-lens/docs/pi-lens-reviewer.md

@@ -1,68 +1,55 @@
 # Reviewer contract
 
-Reviewers run the relevant scenario as a probe and may add a throwaway scenario directory to reproduce a finding through the real host; quote the RPC event or tool result.
-
-Adversarially verify a change before merge and report proven findings.
-
-Assume the implementation's claims are incomplete. Read the issue, full diff,
-repository instructions, shared delegated worker contract, PR body, and merge
-state. Keep the review read-only.
-
-Diff the change from its merge base (`git diff origin/master...HEAD`, or
-`git diff $(git merge-base origin/master HEAD)..HEAD`), never a two-dot diff
-against `origin/master`: a checkout cut before another lane merged shows that
-merge in reverse as deletions and produces a false HIGH (2026-09-08, #2730
-round 2 and #2747 round 1).
-
-Reproduce the build and targeted tests. Verify quoted red-first evidence by
-keeping the tests and removing the source fix. Mutate every new guard and demand
-a red test. Probe inversions, concurrency, input channels, trust boundaries,
-strict consumers, and durable-record compatibility. Repeat the pattern and
-population sweeps. Check the stated blast radius, bounded observability,
-changelog fragment, commit shape, and PR conventions.
-
-Report `CRITICAL`, `HIGH`, `MEDIUM`, `LOW`, then `NITPICK` findings. Give the
-file and line, a concrete failure, evidence, and a suggested fix. Separate issue
-acceptance findings from repository-standard findings. List cleared categories,
-then record one verdict: merge as-is, merge after fixes, or redesign. Never
-merge or silently repair the author's branch. Use short, active, plain prose.
-
-## Tautological tests considered harmful
-
-Check that each regression test reaches the real seam and observes an independent
-effect. Remove or mutate the claimed guard and require the test to fail for the
-intended reason. Flag tests that restate the implementation, assert setup data,
-or swap a real in-process store, sink, coordinator, or registry for a fake.
-
-## Language coverage is a standing attack
-
-For any diff on an LSP, dispatch, cache, runner or tool seam, ask whether the
-rule holds for every entry in `clients/language-registry.ts` or only for the
-language the fixer tested. A rule keyed on `.ts`/tsserver where rust-analyzer,
-pyright or gopls behave the same is a finding (catalog shape 42); probe it
-with one non-TypeScript fixture through the same seam and quote the result.
-
-## Finding shape and disposition
-
-Every finding is written in four moves, in this order: the smallest concrete
-instance (the probe command and its output, or the failing scenario) with the
-expected value beside it; one plain-language sentence saying what is wrong;
-the cause and its cost in prose; the remedy, with options labelled A/B when
-more than one is defensible. Symbols and `file:line` anchor the prose and
-never replace it. Severity is earned by the instance: a CRITICAL or HIGH
-without a reproduced failure scenario is a MEDIUM at most.
-
-The verdict line comes first. After the findings, two fixed sections:
-"Could not verify" (what was blocked and why, never implied green) and
-"Named output" (the structural insight the probes could not close). A Named
-output entry may carry an advisory strength — Strong, Worth exploring, or
-Speculative — judged by the deletion test alone (would removing the shared
-module concentrate complexity back into callers, or merely relocate it); it is
-a triage aid for the orchestrator, never a severity, and never substitutes for
-a reproduced instance on a finding. A verify
-round ends with a disposition table for the previous round's findings —
-`fixed | not fixed | new defect | withdrawn` per finding id — so the
-orchestrator reads outcomes, not prose. Safe deltas (a body sentence, a
-comment, a literal, a doc line) are reported as such and never counted as
-actionable rounds. Borrowed shape: p3bot/library `tasks/review/pre-commit`
-(finding IDs, per-item template, remediation summary), 2026-09-09.
+## Mission
+
+- Read the issue, full merge-base diff, `AGENTS.md`,
+  `docs/pi-lens-subagent.md`, the PR body, and merge state.
+- Keep the branch and worktree read-only.
+- Reproduce the claimed behavior through the production entry point.
+- Report only proven findings; do not repair the author's branch.
+
+## Verification
+
+- Use `git diff origin/master...HEAD` or the merge-base equivalent.
+- Build and run the targeted and required governance suites.
+- Revert or neuter the source fix and verify the red-first test fails.
+- Mutate every new guard, filter, cap, fallback, and lifecycle path.
+- Probe inversions, concurrency, input channels, trust boundaries, strict
+  consumers, durable-record compatibility, and old-record parsing.
+- Repeat the pattern and population sweeps.
+- Check blast radius, bounded observability, changelog, commit, and PR-body
+  requirements.
+- For LSP, dispatch, cache, runner, or tool changes, test one non-TypeScript
+  registry entry through the same seam.
+- On a net-count fold, mutate every predicate the deleted sibling used to
+  back; the fold's own tests were written when two guards existed (#3064 F1,
+  #3065 F3, #3066 F3, #3068 F2).
+
+## Finding format
+
+Order findings: `CRITICAL`, `HIGH`, `MEDIUM`, `LOW`, `NITPICK`.
+
+Each actionable finding contains:
+
+1. Severity and stable id.
+2. File/symbol anchor.
+3. Reproduction command or probe output.
+4. Expected and observed behavior.
+5. Root cause, cost, and concrete remedy.
+6. Issue-acceptance or repository-standard classification.
+
+Severity requires a reproduced failure. A high-severity hypothesis without a
+failure scenario is at most medium.
+
+## Verdict
+
+Start with one verdict: `merge as-is`, `merge after fixes`, or `redesign`.
+Then include:
+
+- `Could not verify`: every blocked or environment-limited check.
+- `Named output`: structural insight not closed by the probes.
+- `Disposition table`: each prior finding as `fixed`, `not fixed`, `new defect`,
+  or `withdrawn (reason)`.
+- Cleared categories and exact-head identity.
+
+Use short, active, plain prose. Never merge, push, commit, or silently repair.

package/node_modules/pi-lens/docs/release-qa-baseline.md

@@ -86,8 +86,30 @@ PR's packed tarball (#2700, the check #2587 was missing). `attw`
 | config-provenance | a project config is LOADED and its provenance is reportable | mcp-stdio | `tools/call` `pilens_effective_config` with the fixture file | result names the fixture's `.pi-lens.json` as a contributing document | tool result text | new — `tests/config/pi-lens-config-schema.test.ts` covers the schema, not the packaged load | — |
 | degradation-visible | a silently-ignored input is RECORDED as a degradation instead of vanishing | mcp-stdio | `tools/call` `pilens_health` with the fixture's project-tier `lsp.enabled` (a global-only setting) loaded | health text carries a `config-ignored` degradation line naming the fixture's `.pi-lens.json` | health tool result text | `clients/degradation-ledger.ts` is the reused machinery; no smoke asserts it end to end | #1605 lane 2 (availability-lifecycle): the degradation-recorded half; #1605 additionally asserts RECOVERY, which this row does not |
 | git-install-loads | a `git:` install of a pushed ref builds and loads in a real pi | git-install | `pi install git:github.com/apmantza/pi-lens@<ref>` then `get_commands` | at least 1 `lens-*` command and at least 4 skills | get_commands response JSON | `scripts/rpc-load-check.mjs` assertion, re-run against the git layout | — |
+| publish-toolchain-pinned | the RELEASE WORKFLOW's publish job runs the npm it pins, and that npm validates the tarball | npm-pack | `npx -y "npm@<packageManager pin>" --version` then `npx -y "npm@<packageManager pin>" publish --dry-run`, both in the scratch export | the pinned invocation reports the pin version AND the dry run exits 0, or npm reports the version is already published after packing; a pinned invocation resolution or registry failure leaves the row UNMEASURED | the two commands with their output | new — `tests/config/release-npm-pin-gate.test.ts` pins release.yml's TEXT; this row is the only thing that RUNS the publish job's toolchain before a tag exists (#2940) | — |
 | tool-smoke-install | every npm/pip entry in the installer registry resolves on a real install | npm-install | `node <export>/scripts/smoke-tools.mjs --install --install-registry --installer-root=<installed>/` (registry `<installed>/dist/clients/installer/index.js`; about 1m 29s cold on this box for 33 entries; harness from export root) | the report shows every npm/pip entry resolved or a named legitimate skip (toolchain absent, declined), and no genuine install failure; a registry-unreachable classification leaves the lane UNMEASURED; requires network access to the npm and pip registries | install-registry JSON report | `classifyInstallOutcome` from the #2661 fixture lanes — this lane sweeps the whole npm/pip registry, where fixture lanes exercise only the entries their fixtures name | — |
 
+## Why `publish-toolchain-pinned` runs on every candidate
+
+#2940: 9183f39c6 moved `release.yml`'s npm pin from a global install to
+`npx -y "npm@<pin>"` and left `npm publish` bare, so the publish job ran Node
+22's bundled npm — no OIDC trusted publishing — and the v4.1.6 run created the
+tag and the GitHub release before the registry answered E404. The `prepare`
+job's `--dry-run` publish could not see it: it ran a different npm.
+
+A `release.yml` change since the last tag is the obvious trigger for this row,
+and it is the one the skill's DIFF-row step (step 2) resolves to. The row runs
+on **every** tree run anyway, because the trigger has a hole: the pin's VALUE
+lives in `package.json`'s `packageManager`, not in the workflow, so a pin bump
+changes exactly the toolchain the publish job uses while leaving `release.yml`
+untouched. A release candidate whose publish path was never executed is the
+whole defect; gating the row on the workflow file would reproduce it one
+`packageManager` bump later.
+
+The row is SKIPPED — never a verdict — under `--from npm:<version>`: there is
+no exported tree to publish, and a dry-run publish fires this package's own
+`prepack`/`prepare`, so it may only ever run in the scratch export.
+
 ## Why `skills-registered` pins the registrar
 
 pi-lens has **two independent skill registrars**: the `pi.skills` manifest, and

package/node_modules/pi-lens/docs/settings.md

@@ -43,10 +43,11 @@ For a single toggle, highest priority first:
 1. **Environment variable**, for the toggles that have one bound (only
    `PI_LENS_NO_CONTEXT_INJECTION` today).
 2. **CLI flag**.
-3. **Nearest project `.pi-lens.json`** that defines the key — for the three
-   project-scoped mutation controls only (`format.enabled`, `autofix.enabled`,
-   `actionableWarnings.autoFix.enabled`). In a monorepo the closest config to the
-   edited file wins.
+3. **Nearest project `.pi-lens.json`** that defines the key — for the
+   project-scoped keys only: the three mutation controls (`format.enabled`,
+   `autofix.enabled`, `actionableWarnings.autoFix.enabled`) and the per-tool
+   switches `tools.<name>.enabled` (`tools.lazy` itself stays global-only). In a
+   monorepo the closest config to the edited file wins.
 4. **Global `~/.pi-lens/config.json`**.
 5. **Built-in default**.
 
@@ -83,6 +84,8 @@ column is the effective behavior when nothing is set.
 | `--lens-actionable-warning-autofix` | `actionableWarnings.autoFix.enabled` | project | **off** |
 | `--lens-actionable-warning-all` | `actionableWarnings.deltaOnly` (`false`) | global | `deltaOnly` **on** (report this turn only) |
 | `--lens-compact-tool-line` | `ui.compactToolLine` | global | **off** (two-row tool rendering) |
+| `--lens-compact-lsp-status` | `ui.compactLspStatus` | global | **off** (footer lists the active server names) |
+| `--lens-hide-lsp-status` | `ui.hideLspStatus` | global | **off** (footer publishes the `pi-lens-lsp` status; outranks `ui.compactLspStatus` when both are set) |
 | `--no-lazy-tools` | `tools.lazy` | global | lazy tools **on** (five situational tools start inactive) |
 | `--no-tool=<name>` | `tools.<name>.enabled` | project | every lens tool **on** |
 | `--lens-turn-end-madge` | `turnEnd.madge.enabled` | global | **off** (madge runs at session start, not per turn) |

package/node_modules/pi-lens/docs/tree-sitter_rules_catalog.md

@@ -218,7 +218,7 @@ Rule source: `rules/tree-sitter-queries/<language>/`.
 | `no-eval` | error | security | eval() detected — security risk, never use eval |
 | `no-jump-in-finally` | warning | bug | return/break/continue/throw in finally overrides the try/catch result and silently swallows exceptions |
 | `self-assignment` | error | reliability | '{{VAR}}' is assigned to itself |
-| `sql-injection` | warning | security | SQL injection risk — use parameterized queries, never interpolate into SQL |
+| `sql-injection` | error | security | SQL injection risk — use parameterized queries, never interpolate into SQL |
 | `switch-case-termination` | error | reliability | Switch case should end with break, return, throw, or continue |
 | `switch-non-case-labels-ts` | error | reliability | switch statements should not contain non-case labels |
 | `ts-command-injection` | error | security | Potential command injection sink — avoid child_process command execution with untrusted input |

package/node_modules/pi-lens/docs/usage.md

@@ -112,6 +112,8 @@ pi --no-opengrep         # Disable the Opengrep security scanner (default-on aux
 pi --no-read-guard       # Disable the read-before-edit behavior monitor
 pi --lens-turn-summary   # Persist a per-turn summary of diagnostics, autofixes, and autoformats
 pi --lens-compact-tool-line   # Render tool results as one compact, theme-aware line (closes #1327)
+pi --lens-compact-lsp-status  # Collapse the footer LSP status to one state glyph instead of the server names (refs #3099)
+pi --lens-hide-lsp-status     # Publish no pi-lens-lsp footer status at all, so a status-rendering host stops showing the key (closes #3099)
 pi --no-lazy-tools       # Keep every pi-lens tool active instead of activating the situational ones on demand
 pi --lens-turn-end-madge # Run the madge circular-dependency check at every turn end, not just at session start
 pi --no-knip             # Disable the knip session-start analyzer

package/node_modules/pi-lens/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "pi-lens",
-	"version": "4.1.6",
+	"version": "4.2.0",
 	"type": "module",
 	"description": "Real-time code feedback for pi — LSP, linters, formatters, type-checking, structural analysis & booboo",
 	"repository": {