@codyswann/lisa 2.240.0 → 2.242.0

package/plugins/lisa/hooks/threshold-ratchet-families.mjs

@@ -0,0 +1,295 @@
+/**
+ * Threshold ratchet — watched file families and value extractors.
+ *
+ * Pure extraction layer: given file contents, produce comparable constraint
+ * maps. No filesystem or git access. See threshold-ratchet.mjs for the CLI
+ * and threshold-ratchet-compare.mjs for the comparison rules.
+ */
+
+/**
+ * File families the ratchet watches. `kind` selects the extractor;
+ * `direction` applies to numeric-leaf kinds ("min" values may only rise,
+ * "max" values may only fall).
+ */
+export const FAMILIES = [
+  {
+    id: "coverage",
+    match: /(^|\/)(vitest|jest)\.thresholds\.json$/,
+    kind: "json-num",
+    direction: "min",
+  },
+  {
+    id: "simplecov",
+    match: /(^|\/)simplecov\.thresholds\.json$/,
+    kind: "json-num",
+    direction: "min",
+  },
+  {
+    id: "e2e",
+    match: /(^|\/)e2e\.thresholds\.json$/,
+    kind: "json-num",
+    direction: "min",
+  },
+  {
+    id: "eslint",
+    match: /(^|\/)eslint\.thresholds\.json$/,
+    kind: "json-num",
+    direction: "max",
+  },
+  {
+    id: "rubocop",
+    match: /(^|\/)rubocop\.thresholds\.yml$/,
+    kind: "rubocop-yaml",
+    direction: "max",
+  },
+  { id: "stryker", match: /(^|\/)stryker\.conf\.json$/, kind: "stryker" },
+  {
+    id: "k6",
+    match: /(^|\/)\.github\/k6\/thresholds\/[^/]+\.json$/,
+    kind: "k6",
+  },
+  {
+    id: "audit-ignore",
+    match: /(^|\/)audit\.ignore\.(config|local)\.json$/,
+    kind: "exemption-list",
+  },
+  {
+    id: "lisa-config",
+    match: /(^|\/)\.lisa\.config\.json$/,
+    kind: "allow-list",
+  },
+];
+
+/**
+ * Find the family a repo-relative path belongs to.
+ * @param {string} relPath Repo-relative path (forward slashes)
+ * @returns {(typeof FAMILIES)[number] | undefined} The matching family, or
+ *   undefined when the path is not a watched gate file
+ */
+export function familyFor(relPath) {
+  return FAMILIES.find(f => f.match.test(relPath));
+}
+
+/**
+ * Safe JSON parse.
+ * @param {string | null | undefined} text JSON text
+ * @returns {unknown | undefined} Parsed value, or undefined on failure
+ */
+export function parseJson(text) {
+  if (typeof text !== "string") return undefined;
+  try {
+    return JSON.parse(text);
+  } catch {
+    return undefined;
+  }
+}
+
+/**
+ * Walk a JSON object and collect numeric leaves as dotted-path constraints.
+ * Keys starting with "_" (e.g. `_comment`) are documentation, not thresholds.
+ * @param {unknown} node Parsed JSON value
+ * @param {"min"|"max"} direction Ratchet direction for every leaf
+ * @param {string} [prefix] Dotted path accumulated so far
+ * @returns {Map<string, { value: number, direction: "min"|"max" }>} Dotted
+ *   path → numeric constraint for every finite numeric leaf
+ */
+export function extractNumericLeaves(node, direction, prefix = "") {
+  const out = new Map();
+  if (node === null || typeof node !== "object" || Array.isArray(node)) {
+    return out;
+  }
+  for (const [key, value] of Object.entries(node)) {
+    if (key.startsWith("_")) continue;
+    const p = prefix ? `${prefix}.${key}` : key;
+    if (typeof value === "number" && Number.isFinite(value)) {
+      out.set(p, { value, direction });
+    } else if (value && typeof value === "object" && !Array.isArray(value)) {
+      for (const [cp, c] of extractNumericLeaves(value, direction, p)) {
+        out.set(cp, c);
+      }
+    }
+  }
+  return out;
+}
+
+/**
+ * Minimal parser for rubocop.thresholds.yml — a two-level document of
+ * `Section:` headers with indented `Key: <number>` scalars (comments and
+ * blank lines ignored). Deliberately NOT a general YAML parser: the file is
+ * Lisa-authored with this exact shape, and hand-parsing keeps the gate
+ * dependency-free with no backtracking-prone regexes.
+ * @param {string} text File contents
+ * @param {"min"|"max"} direction Ratchet direction for every scalar
+ * @returns {Map<string, { value: number, direction: "min"|"max" }>} Dotted
+ *   `Section.Key` path → numeric constraint
+ */
+export function extractRubocopThresholds(text, direction) {
+  const out = new Map();
+  const state = { section: "" };
+  for (const rawLine of text.split("\n")) {
+    const hash = rawLine.indexOf("#");
+    const line = (hash >= 0 ? rawLine.slice(0, hash) : rawLine).trimEnd();
+    if (!line.trim()) continue;
+    const indented = line.startsWith(" ") || line.startsWith("\t");
+    if (!indented && line.endsWith(":")) {
+      state.section = line.slice(0, -1).trim();
+      continue;
+    }
+    const colon = line.indexOf(":");
+    if (!indented || colon < 0 || !state.section) continue;
+    const key = line.slice(0, colon).trim();
+    const rawValue = line.slice(colon + 1).trim();
+    // Number("") is 0, so an empty value (e.g. a nested `Exclude:` list
+    // header) must be skipped, not recorded as a zero threshold.
+    const value = Number(rawValue);
+    if (key && rawValue !== "" && Number.isFinite(value)) {
+      out.set(`${state.section}.${key}`, { value, direction });
+    }
+  }
+  return out;
+}
+
+/**
+ * Extract the gating constraint from stryker.conf.json: only
+ * `thresholds.break` fails a run (`high`/`low` are reporting bands).
+ * @param {unknown} conf Parsed stryker.conf.json
+ * @returns {Map<string, { value: number, direction: "min"|"max" }>} The
+ *   `thresholds.break` constraint when present, otherwise an empty map
+ */
+export function extractStrykerConstraints(conf) {
+  const out = new Map();
+  const breakValue = conf?.thresholds?.break;
+  if (typeof breakValue === "number" && Number.isFinite(breakValue)) {
+    out.set("thresholds.break", { value: breakValue, direction: "min" });
+  }
+  return out;
+}
+
+/**
+ * Split a stryker `mutate` array into positive globs and negations.
+ * @param {unknown} conf Parsed stryker.conf.json
+ * @returns {{ positives: Set<string>, negations: Set<string> }} Globs that
+ *   include files vs. `!`-prefixed globs that exclude them
+ */
+export function extractStrykerMutate(conf) {
+  const positives = new Set();
+  const negations = new Set();
+  const mutate = Array.isArray(conf?.mutate) ? conf.mutate : [];
+  for (const glob of mutate) {
+    if (typeof glob !== "string") continue;
+    if (glob.startsWith("!")) negations.add(glob);
+    else positives.add(glob);
+  }
+  return { positives, negations };
+}
+
+/**
+ * Parse one k6 threshold expression (`p(95)<1000`, `rate>=0.99`, …) into a
+ * ratchet constraint. Upper bounds (`<`, `<=`) may only decrease; lower
+ * bounds (`>`, `>=`) may only increase. Hand-parsed — no regex.
+ * @param {string} expr Threshold expression
+ * @returns {{ value: number, direction: "min"|"max" } | undefined} The bound
+ *   as a constraint, or undefined when the expression has no numeric bound
+ */
+export function parseK6Expression(expr) {
+  for (const op of ["<=", ">=", "<", ">"]) {
+    const idx = expr.indexOf(op);
+    if (idx < 0) continue;
+    const value = Number(expr.slice(idx + op.length).trim());
+    if (!Number.isFinite(value)) return undefined;
+    return { value, direction: op.startsWith("<") ? "max" : "min" };
+  }
+  return undefined;
+}
+
+/**
+ * Extract constraints from a k6 thresholds file. Each metric contributes its
+ * expression bound(s) plus (when present) `abortOnFail` boolean constraints —
+ * turning abortOnFail off makes the gate advisory, which is a weakening.
+ * Handles every documented k6 shape: a bare expression string, a single long
+ * form `{ threshold, abortOnFail }` object, and arrays mixing both.
+ * @param {unknown} conf Parsed k6 thresholds JSON
+ * @returns {{
+ *   numeric: Map<string, { value: number, direction: "min"|"max" }>,
+ *   booleans: Map<string, boolean>,
+ * }} Numeric bounds keyed by `<metric>.threshold[i]` and abortOnFail flags
+ *   keyed by `<metric>.abortOnFail` (`[i]`-suffixed for array items)
+ */
+export function extractK6Constraints(conf) {
+  const numeric = new Map();
+  const booleans = new Map();
+  const thresholds = conf?.thresholds;
+  if (thresholds && typeof thresholds === "object") {
+    for (const [metric, spec] of Object.entries(thresholds)) {
+      const isArray = Array.isArray(spec);
+      const items = isArray ? spec : [spec];
+      items.forEach((item, i) => {
+        const expr =
+          typeof item === "string"
+            ? item
+            : item && typeof item === "object" && !Array.isArray(item)
+              ? item.threshold
+              : undefined;
+        if (typeof expr === "string") {
+          const c = parseK6Expression(expr);
+          if (c) numeric.set(`${metric}.threshold[${i}]`, c);
+        }
+        if (
+          item &&
+          typeof item === "object" &&
+          !Array.isArray(item) &&
+          typeof item.abortOnFail === "boolean"
+        ) {
+          booleans.set(
+            isArray ? `${metric}.abortOnFail[${i}]` : `${metric}.abortOnFail`,
+            item.abortOnFail
+          );
+        }
+      });
+    }
+  }
+  return { numeric, booleans };
+}
+
+/**
+ * Flatten an exemption file (audit ignore list) into a set of entry tokens.
+ * Arrays contribute their string items; objects contribute their keys.
+ * @param {unknown} conf Parsed JSON
+ * @returns {Set<string>} One token per exemption entry
+ */
+export function extractExemptionEntries(conf) {
+  const out = new Set();
+  if (Array.isArray(conf)) {
+    for (const item of conf) {
+      if (typeof item === "string") out.add(item);
+      else if (item && typeof item === "object") out.add(JSON.stringify(item));
+    }
+  } else if (conf && typeof conf === "object") {
+    for (const [key, value] of Object.entries(conf)) {
+      if (Array.isArray(value)) {
+        for (const item of value) {
+          out.add(
+            `${key}:${typeof item === "string" ? item : JSON.stringify(item)}`
+          );
+        }
+      } else {
+        out.add(key);
+      }
+    }
+  }
+  return out;
+}
+
+/**
+ * Extract thresholdRatchet.allow entries from parsed .lisa.config.json.
+ * @param {unknown} config Parsed config (may be undefined)
+ * @returns {Array<{ file: string, key: string, reason?: string }>} The
+ *   well-formed allow entries; malformed entries are dropped
+ */
+export function extractAllowEntries(config) {
+  const raw = config?.thresholdRatchet?.allow;
+  if (!Array.isArray(raw)) return [];
+  return raw.filter(
+    e => e && typeof e.file === "string" && typeof e.key === "string"
+  );
+}

package/plugins/lisa/hooks/threshold-ratchet.mjs

@@ -0,0 +1,213 @@
+#!/usr/bin/env node
+/**
+ * Threshold ratchet gate — quality thresholds may tighten, never weaken.
+ *
+ * Deterministic comparator shared by three enforcement layers:
+ *   1. Agent-time soft block: PostToolUse hook via threshold-ratchet.sh
+ *      (`--hook`, exit 2 on weakening so the agent gets actionable feedback).
+ *   2. Pre-commit backstop: husky / lefthook (`--staged`, exit 1).
+ *   3. CI gate: reusable quality workflows (`--base <ref>`, exit 1),
+ *      comparing against the merge-base so nothing weakened lands in a PR.
+ *
+ * Tier 1 — designed tunables: vitest/jest/simplecov/e2e thresholds
+ * (minimums) and eslint/rubocop thresholds (maximums). Tier 2 — stryker's
+ * break score and k6 expression bounds. Tier 3 — exemption additions
+ * (audit-ignore entries, stryker mutate exclusions, thresholdRatchet.allow
+ * entries) which weaken a gate without touching a number.
+ *
+ * Human override: `.lisa.config.json` → `thresholdRatchet.allow` entries
+ * ({ file, key, reason }). Honored ONLY from the baseline side (HEAD /
+ * merge-base), never from the change under review — an agent cannot grant
+ * itself an exception in the same change that weakens a gate. `key: "*"`
+ * allows every key in the file.
+ *
+ * Extraction lives in threshold-ratchet-families.mjs; comparison rules in
+ * threshold-ratchet-compare.mjs. Zero dependencies.
+ */
+import { execFileSync } from "node:child_process";
+import * as fs from "node:fs";
+import * as path from "node:path";
+import { fileURLToPath } from "node:url";
+import {
+  extractAllowEntries,
+  familyFor,
+  parseJson,
+} from "./threshold-ratchet-families.mjs";
+import {
+  applyAllowList,
+  compareFile,
+  formatReport,
+} from "./threshold-ratchet-compare.mjs";
+
+/**
+ * Standard git locations, checked in order so the executable comes from a
+ * fixed, unwriteable directory rather than a PATH lookup. The bare "git"
+ * fallback keeps unusual layouts (e.g. Windows git-bash) working.
+ */
+const GIT_LOCATIONS = [
+  "/usr/bin/git",
+  "/usr/local/bin/git",
+  "/opt/homebrew/bin/git",
+];
+const GIT = GIT_LOCATIONS.find(candidate => fs.existsSync(candidate)) ?? "git";
+
+/** Git flag shared by every changed-file listing. */
+const NAME_ONLY = "--name-only";
+
+/**
+ * Run git, returning stdout or null on any failure.
+ * @param {string[]} args Git arguments
+ * @param {string} [cwd] Working directory
+ * @returns {string | null} Captured stdout, or null when git failed
+ */
+function git(args, cwd) {
+  try {
+    return execFileSync(GIT, args, {
+      cwd,
+      encoding: "utf-8",
+      stdio: ["ignore", "pipe", "ignore"],
+    });
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Resolve mode-specific candidate files and content readers.
+ * @param {"hook"|"staged"|"base"} mode Comparison mode
+ * @param {string} root Repo root
+ * @param {string | undefined} baseRef Base ref (base mode only)
+ * @param {string[] | undefined} onlyFiles Restrict to these repo-relative
+ *   paths (hook mode with a known edited file)
+ * @returns {{ files: string[], baselineRef: string, readCurrent: (f: string) => string | null } | null}
+ *   The comparison plan, or null when git state can't support the mode
+ */
+function resolvePlan(mode, root, baseRef, onlyFiles) {
+  if (mode === "staged") {
+    const diff = git(["diff", "--cached", NAME_ONLY], root);
+    if (diff === null) return null;
+    return {
+      files: diff.split("\n").filter(Boolean),
+      baselineRef: "HEAD",
+      readCurrent: f => git(["show", `:${f}`], root),
+    };
+  }
+  if (mode === "base") {
+    if (!baseRef) return null;
+    const mergeBase = git(["merge-base", baseRef, "HEAD"], root)?.trim();
+    if (!mergeBase) return null;
+    const diff = git(["diff", NAME_ONLY, mergeBase, "HEAD"], root);
+    if (diff === null) return null;
+    return {
+      files: diff.split("\n").filter(Boolean),
+      baselineRef: mergeBase,
+      readCurrent: f => git(["show", `HEAD:${f}`], root),
+    };
+  }
+  const diff = git(["diff", NAME_ONLY, "HEAD"], root);
+  if (diff === null) return null;
+  return {
+    files: onlyFiles ?? diff.split("\n").filter(Boolean),
+    baselineRef: "HEAD",
+    readCurrent: f => {
+      try {
+        return fs.readFileSync(path.join(root, f), "utf-8");
+      } catch {
+        return null;
+      }
+    },
+  };
+}
+
+/**
+ * Run the ratchet for a mode, print the report, and return the exit code.
+ * @param {"hook"|"staged"|"base"} mode Comparison mode
+ * @param {string | undefined} [baseRef] Base ref (base mode only)
+ * @param {string[] | undefined} [onlyFiles] Restrict to these paths (hook mode)
+ * @returns {number} Process exit code (2 for hook mode, 1 otherwise; 0 clean)
+ */
+function run(mode, baseRef, onlyFiles) {
+  const root = git(["rev-parse", "--show-toplevel"])?.trim();
+  if (!root) return 0;
+  const plan = resolvePlan(mode, root, baseRef, onlyFiles);
+  if (!plan) return 0;
+
+  const watched = plan.files.filter(f => familyFor(f));
+  if (watched.length === 0) return 0;
+
+  const findings = watched.flatMap(f =>
+    compareFile(
+      f,
+      git(["show", `${plan.baselineRef}:${f}`], root),
+      plan.readCurrent(f)
+    )
+  );
+  if (findings.length === 0) return 0;
+
+  const baselineConfig = parseJson(
+    git(["show", `${plan.baselineRef}:.lisa.config.json`], root)
+  );
+  const { blocked, allowed } = applyAllowList(
+    findings,
+    extractAllowEntries(baselineConfig)
+  );
+  for (const finding of allowed) {
+    process.stdout.write(
+      `threshold-ratchet: allowed by .lisa.config.json exception — ${finding.message}\n`
+    );
+  }
+  if (blocked.length === 0) return 0;
+  process.stderr.write(`${formatReport(blocked)}\n`);
+  return mode === "hook" ? 2 : 1;
+}
+
+/**
+ * Handle `--hook` mode: parse the tool-use event from stdin and scope the
+ * check to the edited file (Edit/Write/NotebookEdit) or every changed
+ * watched file (Bash).
+ * @returns {number} Process exit code
+ */
+function runHookMode() {
+  const state = { stdin: "" };
+  try {
+    state.stdin = fs.readFileSync(0, "utf-8");
+  } catch {
+    return 0;
+  }
+  const input = parseJson(state.stdin);
+  if (!input || typeof input !== "object") return 0;
+  if (input.tool_name === "Bash") return run("hook");
+  if (!["Edit", "Write", "NotebookEdit"].includes(input.tool_name)) return 0;
+  const filePath = input.tool_input?.file_path;
+  if (typeof filePath !== "string") return 0;
+  const root = git(["rev-parse", "--show-toplevel"])?.trim();
+  if (!root) return 0;
+  const rel = path
+    .relative(root, path.resolve(filePath))
+    .split(path.sep)
+    .join("/");
+  if (rel.startsWith("..") || !familyFor(rel)) return 0;
+  return run("hook", undefined, [rel]);
+}
+
+/**
+ * CLI entrypoint.
+ * @returns {number} Process exit code
+ */
+function main() {
+  const args = process.argv.slice(2);
+  if (args[0] === "--staged") return run("staged");
+  if (args[0] === "--base") return run("base", args[1]);
+  if (args[0] === "--hook") return runHookMode();
+  process.stderr.write(
+    "usage: threshold-ratchet.mjs --hook | --staged | --base <ref>\n"
+  );
+  return 0;
+}
+
+const isDirectRun =
+  process.argv[1] &&
+  fileURLToPath(import.meta.url) === path.resolve(process.argv[1]);
+if (isDirectRun) {
+  process.exit(main());
+}

package/plugins/lisa/hooks/threshold-ratchet.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+# PostToolUse hook for Edit|Write|NotebookEdit|Bash: the threshold ratchet.
+# Quality thresholds (coverage minimums, complexity maximums, mutation break
+# score, e2e route floors, k6 bounds, audit-ignore lists) may tighten but never
+# weaken. The deterministic comparator lives in threshold-ratchet.mjs and is
+# shared with the pre-commit (husky/lefthook --staged) and CI (--base) layers,
+# so this hook is fast feedback — not the only line of defense.
+#
+# Exit 2 on weakening (soft block: the agent gets the report on stderr and can
+# fix the code or escalate to a human). Every infrastructure gap — no node, no
+# git repo, unreadable stdin — exits 0: the CI layer still guarantees the gate,
+# and a broken hook must never wedge an agent session.
+set -euo pipefail
+
+input="$(cat)"
+
+command -v node >/dev/null 2>&1 || exit 0
+git rev-parse --is-inside-work-tree >/dev/null 2>&1 || exit 0
+
+script_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+printf '%s' "$input" | node "$script_dir/threshold-ratchet.mjs" --hook

package/plugins/lisa/rules/eager/claim-archaeology.md

@@ -0,0 +1,37 @@
+# Claim-Time Archaeology (load-bearing)
+
+Lisa lifecycles are one-way — a done issue never reopens, so a residual failure comes back as a **new** issue with no visible link to the issue that shipped it. Archaeology recovers that link at claim time: the claiming agent learns it is working on round 2 of a past failure, and what specifically went wrong the first time.
+
+**One vendor-neutral contract, cited by every build-intake arm** (the `leaf-only-lifecycle` / `repo-scope-split` / `rejection-detection` precedent: one shared slug, never three divergent implementations).
+
+## When it runs
+
+In build-intake step 3b, **AFTER the rejection-detection classification and BEFORE the relabel/transition** `$READY → $CLAIMED`. Rejection detection runs first; its classification is an **input** to archaeology — a detected `rejection-reclaim` passes straight through, never re-derived.
+
+## Classify the claimed item
+
+Return exactly one of:
+
+- **`rejection-reclaim`** — taken directly from the `rejection-detection` result. Reuse it; do not re-derive.
+- **`retry-of-done-issue`** — an ancestry signal names a closed done issue whose shipped work this issue exists to fix.
+- **`fresh`** — no ancestor found, signals weak/inconclusive, budget exhausted, or the pass errored. The default and the safe degrade.
+
+## Ancestry signals (summary — full bindings in the reference body)
+
+1. **Tracker metadata** — the typed relations the read skills already parse (Blocks / Blocked by / Relates to / Duplicates / Cloned from, `closingIssuesReferences`, cross-references).
+2. **Text similarity** — tracker search primitives over recently-closed issues touching the same implicated files, ranked by title/label overlap. Lexical only; no embedding machinery exists.
+3. **Git ancestry** — deterministic `git log --follow` / `git blame` / merge-commit queries yielding a parseable `{file, sha, pr, date}` result. Never delegate this to the prose-report `git-history-analyzer` agent.
+
+## Learning-loop exclusion (scan-side)
+
+An artifact this flow produced is **never** an ancestor. Exclude anything carrying `[lisa-learning-drop]`, `[lisa-learning-pr]`, `[lisa-learning-upstream-handoff]`, `[lisa-rejection-candidate]`, or `[lisa-archaeology-candidate]` markers, or the `learning:needs-triage` label.
+
+## Cost budget — never block the claim
+
+The pass runs inside a hard budget: `.lisa.config.json` `archaeology.maxSteps` (default **8** tracker/git queries; optional `archaeology.maxSeconds`). Budget exhausted, weak signals, or an exception → classify `fresh` and proceed. Archaeology is a bonus layered on the claim; it **never blocks the claim**. Exceeding the budget degrades to `fresh` — a normal outcome, not an error.
+
+## On `retry-of-done-issue`
+
+Reconstruct what the ancestor's PR shipped (merged PR, review threads, evidence comments) and derive **ONE** candidate learning citing the **delta** between what was done and what this issue proves was needed — routed to `lisa-persist-learning` exactly like the rejection-reflection path. Fallback when that skill is absent: a comment with a visible prose line plus `<!-- [lisa-archaeology-candidate] key=<issue>::<ancestor> -->` (marker-dedupe; re-claims produce no duplicate). `fresh` → no candidate, zero comments.
+
+Full contract (signal bindings, classification table, candidate derivation, budget mechanics): [reference/claim-archaeology.md](../reference/claim-archaeology.md).