genesis-compiler 1.10.12 → 1.10.14

package/README.md

@@ -736,9 +736,13 @@ The eight task types are deliberately explicit:
 - `describe` creates or refreshes both the non-technical Blueprint and the
   complete useful Program from the current codebase, optional user guidance,
   and selected Stack guidance where it affects accurate explanation.
-- `review` is read-only. It asks the agent to compare Blueprint, code, Program,
-  and tests, identify contradictions and gaps, and distinguish evidence from
-  inference.
+- `review` asks the agent to compare Blueprint, code, Program, and tests,
+  identify contradictions and gaps, and distinguish evidence from inference.
+  It judges the tests too: the relevant ones run and pass, with what ran and
+  what was not verified reported; changed behavior has tests; no test was
+  weakened, skipped, deleted or loosened to pass; and the changed tests stay
+  fast against coarse sanity limits. It edits only tests, writing the missing
+  ones or reporting those it cannot.
 
 Prompt generation is read-only. Missing external resources are reported in
 the prompt but do not prevent its generation. Genesis never invents values.
@@ -1074,7 +1078,10 @@ and diagnostics for missing, invalid, outdated, customized, or unmanaged skills.
 Synchronization returns changed paths, preservation diagnostics, and the resulting
 `inspection`. Hosts own source-write exclusion and assistant context lifecycle.
 `projectSessionContext()` composes stable project guidance with one explicitly
-supplied host-driver result. `projectTurnContext()` invokes that same optional
+supplied host-driver result. A project file Genesis cannot read does not stop
+it: the context names the file, the line and the text Genesis found there, so
+the agent can repair it. That includes an Agent Skill source and an agent
+integration file such as `.claude/settings.json`. `projectTurnContext()` invokes that same optional
 driver for the separate untrusted turn lane without accepting user-message
 text.
 `getContext()` resolves source paths to the Program modules that cite them,

package/docs/prompt-integration.md

@@ -161,8 +161,18 @@ before selection. For an existing uninitialized codebase it owns
 the adoption recommendation as part of the same portable prompt. A host sends this prompt on the
 first chat turn; it does not recreate the questions or Stack catalog.
 `describe` creates or refreshes the complete Blueprint and useful Program in one
-agent task. The host owns the agent and its interaction. Genesis owns only
-transparent prompt construction and deterministic project operations.
+agent task. `review` is the generic review contract a host gives an agent after
+a piece of work. Besides comparing Blueprint, code, Program and tests, it has
+the agent judge the tests: the relevant ones run and pass, with what ran and
+what was not verified reported; changed behavior has new or updated tests,
+written in the review or reported when they cannot be; a test weakened,
+skipped, deleted or loosened to pass is a finding; and the changed tests stay
+fast against coarse sanity limits (a suite never runs for hours, a test needing
+minutes is a smell, a test database is ready in seconds), never per-test
+budgets. It points at the project skill's verification section and the sanity
+checks guide rather than restating them, and edits only tests. The host owns
+the agent and its interaction. Genesis owns only transparent prompt
+construction and deterministic project operations.
 
 ## Instruction sources
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "genesis-compiler",
-  "version": "1.10.12",
+  "version": "1.10.14",
   "type": "module",
   "description": "An agent-independent prompt, multi-language code-index, cleanup, and verification companion with project agent guidance.",
   "repository": {

package/plugins/genesis/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "genesis",
-  "version": "1.10.12",
+  "version": "1.10.14",
   "description": "Makes Codex aware of optional Genesis adoption for existing projects.",
   "author": {
     "name": "Mobily Enterprises"

package/prompts/review.txt

@@ -1,6 +1,5 @@
 Read `.agents/skills/genesis-project/SKILL.md` and
-`.agents/skills/genesis-program/SKILL.md` completely, then perform a read-only
-review.
+`.agents/skills/genesis-program/SKILL.md` completely, then review the work.
 
 Compare Blueprint intent, observable implementation, tests, Program public
 contracts, informational implementation maps, and recorded verification in both
@@ -8,6 +7,21 @@ directions. Identify missing or contradictory behavior, misleading explanation,
 unnecessarily difficult public shape, and verification gaps. Distinguish direct
 evidence from inference and state questions that require human intent.
 
+Judge the tests as part of the work, with the project skill's "Verify and
+report" section and the test sections of the sanity checks guide:
+- Run the tests relevant to the change, through the Stack's verification
+  commands or focused checks; a failure is a finding. Report what actually ran
+  and anything not verified; never report an unrun check as passing.
+- Changed behavior has tests, new or updated, that would fail without it. Write
+  the missing ones as part of this review; when that is not possible, report
+  each with the behavior it should prove.
+- A test weakened, skipped, deleted or loosened just to make things pass is a
+  finding, reported like a product defect.
+- Tests stay fast. Look at what the changed tests cost and where their time
+  goes. Coarse sanity limits, never per-test budgets: a whole suite never runs
+  for hours, one test that needs minutes is a smell, and a test database is
+  ready in seconds.
+
 Report what makes requests or tests slower than their work, citing a count or
 timing where one exists and the code path otherwise: a list or batch whose
 query count grows with its rows, an access check whose cost grows with stored
@@ -21,7 +35,9 @@ access decided from ids the client sends rather than the trusted context.
 Judge each against the matching section and outer limit of
 `.agents/skills/genesis-project/references/sanity-checks.md`.
 
-Do not edit any file. Return a concise report ordered by impact with exact file
-references, the evidence inspected, and the smallest sensible correction.
+Edit only tests: add or update the tests the changed behavior lacks, and change
+no product code, Blueprint, Program, Stack or other file. Return a concise
+report ordered by impact with exact file references, the evidence inspected,
+the tests run and written, and the smallest sensible correction.
 
 Review `genesis/subsystems.md` responsibilities, Program membership, and data ownership/use against source and schema. Report missing, stale, or conflicting associations without editing them.

package/skills/genesis-program/SKILL.md

@@ -71,14 +71,14 @@ Program module.
 ## Task boundaries
 
 The caller determines whether this is initial description, complete Program
-refresh, focused post-change reconciliation, Blueprint-only work, or read-only
-review. Respect the caller's edit boundary. Report ambiguity rather than
+refresh, focused post-change reconciliation, Blueprint-only work, or review.
+Respect the caller's edit boundary. Report ambiguity rather than
 inventing intent.
 
 ## Subsystems
 
 Maintain `genesis/subsystems.md` alongside Program in description and refresh
-work. Blueprint-only work does not change the map; review remains read-only.
+work. Blueprint-only work and review do not change the map.
 The map is the authored association authority, but responsibilities and table
 references remain fallible explanations of application source and schema.
 

package/src/index/agent-skills.js

@@ -315,6 +315,12 @@ export async function projectSkillPlan(stack, { projectRoot } = {}) {
   if (projectRoot) {
     let manifest;
     try { manifest = JSON.parse(await readFile(path.join(projectRoot, 'package.json'), 'utf8')); } catch (error) {
+      // The manifest says whether the project declares its own genesis-compiler skills.
+      if (error instanceof SyntaxError) {
+        throw new GenesisError('AGENT_SKILL_SOURCE_INVALID',
+          `package.json is not valid JSON, so the project's genesis-compiler skills cannot be located: ${error.message}.`,
+          { path: 'package.json' });
+      }
       if (!['ENOENT', 'ENOTDIR'].includes(error?.code)) throw error;
     }
     if (manifest?.name === 'genesis-compiler') {

package/src/index/session-context.js

@@ -1,14 +1,30 @@
 import { readCollaboration } from './collaboration.js';
+import { inspectAgentIntegrations } from './agent-integrations.js';
 import { inspectProjectSkills, renderAgentSkillCatalog } from './agent-skills.js';
 import { GENESIS_CONTRACTS } from './contracts.js';
 import { readEngineering, readEngineeringBaseline } from './engineering.js';
 import { asDiagnostic } from './errors.js';
 import { gitContext } from './git.js';
+import { STACK_PATH } from './paths.js';
 import { inspectProjectAtRoot } from './project-inspection.js';
 import { readStack } from './stack.js';
 import { listStackCatalogPieces } from './stack-catalog.js';
 import { composedContext, hostContextContribution } from './host-context.js';
 
+/**
+ * A diagnostic as the agent needs it to repair the file: the file, the line and the text found there,
+ * when Genesis knows them. Stack section parsers number lines within their section.
+ */
+function diagnosticText({ message, details }) {
+  const file = typeof details?.path === 'string' ? details.path : '';
+  const line = Number.isSafeInteger(details?.line)
+    ? ` (line ${details.line}${file === STACK_PATH ? ' of its section' : ''})`
+    : '';
+  const where = file && (line || !message.includes(file)) ? `${file}${line}: ` : '';
+  const found = typeof details?.observed === 'string' ? ` Found: ${JSON.stringify(details.observed)}.` : '';
+  return `${where}${message}${found}`;
+}
+
 async function optionalStack(projectRoot, stackPackages) {
   try { return await readStack(projectRoot, { stackPackages }); } catch { return null; }
 }
@@ -68,11 +84,16 @@ export async function projectSessionContext({
     inspectProjectAtRoot(root, { stackPackages, count: false }),
   ]);
   const availableComponents = await optionalStackComponentIds(root, stack, stackPackages);
-  const skills = stack && inspection.projectFormat.status === 'current'
-    ? await inspectProjectSkills({ projectRoot: root, stack }).catch((error) => ({
-        diagnostics: [asDiagnostic(error)],
-      }))
-    : { diagnostics: [] };
+  const current = inspection.projectFormat.status === 'current';
+  const [skills, integrations] = await Promise.all([
+    stack && current
+      ? inspectProjectSkills({ projectRoot: root, stack }).catch((error) => ({
+          diagnostics: [asDiagnostic(error)],
+        }))
+      : { diagnostics: [] },
+    // Only an integration file Genesis cannot read is reported; a missing or outdated one is synchronization work.
+    current ? inspectAgentIntegrations({ projectRoot: root }).then(() => [], (error) => [asDiagnostic(error)]) : [],
+  ]);
   const selected = stack?.components.map(({ id }) => id) || [];
   const stackStatus = stack
     ? (selected.length > 0 ? selected.join(', ') : 'none')
@@ -85,7 +106,7 @@ export async function projectSessionContext({
     ] : inspection.state === 'adoption' ? [
       '- Existing project content needs description and configuration. Ask what this project does and what the user wants to run, using answers already supplied. Inspect its implementation to work backwards into Blueprint, Stack and Program. Do not offer seeds or replace existing source/history. Run `genesis adopt` and follow its prompt; initialized Genesis files already express the user’s decision to use Genesis.',
     ] : inspection.state === 'attention' ? [
-      `- Genesis needs attention: ${inspection.diagnostics.map(({ message }) => message).join(' ')} Next action: ${inspection.nextAction}. Preserve source and explain the specific issue before the relevant migration or repair. Do not offer seeds.`,
+      `- Genesis needs attention: ${inspection.diagnostics.map(diagnosticText).join(' ')} Next action: ${inspection.nextAction}. Preserve source and explain the specific issue before the relevant migration or repair. Do not offer seeds.`,
     ] : []),
     '- Opening or resuming a session only reads project metadata and content paths. Do not run verification, tests, builds, dependency installation, or database preparation just because a session started. Run relevant checks when the requested work calls for them.',
     '- Read `genesis/blueprint.md`, `genesis/collaboration.md`, `genesis/engineering.md`, and `genesis/stack.md` for product intent, collaboration approach, engineering approach, and selected technology.',
@@ -95,8 +116,9 @@ export async function projectSessionContext({
     '- When the user names an unselected technology that exactly matches this catalog, run `genesis stack list`, ask whether to add and prepare it, and wait for confirmation. If confirmed, run `genesis stack add <piece...>` and follow its returned preparation prompt in the same task; then use `genesis context` and any installed technology skill. Never infer dependency commands from a component id. If declined or unmatched, continue through authoritative technology documentation without inventing a component.',
     '- Project-owned operation sections in `genesis/stack.md` are the durable application contract. Make the implementation satisfy them or update a complete section to match evidenced reality; Genesis executes only Verification.',
     '- Use the relevant project Agent Skills below `.agents/skills/`.',
-    ...skills.diagnostics.map(({ message }) => `- Agent Skill maintenance: ${message}`),
+    ...skills.diagnostics.map((diagnostic) => `- Agent Skill maintenance: ${diagnosticText(diagnostic)}`),
     ...(skills.diagnostics.length > 0 ? ['- Skill inspection is read-only. Synchronization is an explicit source change; follow the current task authorization and host write boundary before running it.'] : []),
+    ...integrations.map((diagnostic) => `- Agent integration maintenance: ${diagnosticText(diagnostic)}`),
     '- After locating source, run `genesis context <path...>`; before adding a helper or public operation, run `genesis index <name-or-path...>` and reuse an existing owner.',
     '- If genesis/subsystems.md is missing or empty in an existing explained application, author it from source, schema and existing Program before implementation. This is ordinary explanatory adoption, not a startup blocker; do not run tests or application commands merely to create the map.',
     '- Read genesis/subsystems.md for declared subsystem responsibilities, Program membership, and data ownership/use. Maintain it in the same implementation turn when those facts change; reuse existing boundaries, verify source/schema, and require one Source Directory per subsystem and assign each Program module once; Program folder names do not assign ownership.',
@@ -118,7 +140,7 @@ export async function projectSessionContext({
     ...(skills.skills?.length ? ['', 'PROJECT AGENT SKILLS', '', renderAgentSkillCatalog(skills.skills)] : []),
     ...(hostContext ? ['', 'HOST CONTEXT', '', hostContext] : []),
   ].join('\n');
-  return { ...composedContext(GENESIS_CONTRACTS.sessionContext, output), warnings: skills.diagnostics };
+  return { ...composedContext(GENESIS_CONTRACTS.sessionContext, output), warnings: [...skills.diagnostics, ...integrations] };
 }
 
 export { projectTurnContext } from './host-context.js';