Resume project checkpoints and require SI intent review

adapters/chatgpt/metadata/chatgpt-adapter.json

@@ -20,9 +20,9 @@
   "behavioral_contract": "source_preserved_not_installed_behavior_proof",
   "source_manifest_sha256": "0797e9d318206258af21b0152dcb61949c2478fab2837ecdb9bcc67e67f66c89",
   "source_file_sha256": {
-    "SKILL.md": "87eb73b0792238f5e909b20bee3ddfc3c0740ee40471874fc76e877dd765d2e5",
+    "SKILL.md": "9b5657238bedeb78e917f7736592c08311464212e385a0303222d716be0e1e5c",
     "VERSION": "326dd5d8641f399990de42def50ac29eb4c9222601c0a0a3fc3c947dd3bec65d",
-    "agents/openai.yaml": "5090139b222271205759e9c1cd34fbe154a0aae91221cb9fea47e7ecaf29e56c",
+    "agents/openai.yaml": "e534ca4016e2bb2b748aec5f0a57b4e16387db63ee497cc871f5d99082d2d5fb",
     "assets/icon.svg": "bbd2fc7be4ae81f8ff16bd26e26eab6d8547ef3ba1d6d7e49ba970c9a28fc77f",
     "evals/evals.json": "ae07d8148293f31cbf25ac2c3034e98963067e38bfa974dbca13e9126008d563",
     "lanes/si.execution.json": "0277bcf9385a1c4059f67146e24fefb2b541ca6051470350acd582b9ca6c4bff",
@@ -32,7 +32,7 @@
     "references/actual-intent-alignment.md": "3bc1fac79f7fed56c098c5dd01ce7c5e49a973c9a2455606f1ff7c495373a9ad",
     "references/architecture-reuse.md": "5d1c4d09ca8e7407d19c7fc9d68587b2ea0b61be61e5c3448890af2c8766138d",
     "references/continuity-and-impact.md": "78f93ccaaed14e58544c5a0c78b6bf933922b4c576a597b13d63ac74d0de53e0",
-    "references/durable-progress-and-recovery.md": "1457eed72572aa6e0f34ab6773b81735e3f1ee22e5427e44a238619b034a609d",
+    "references/durable-progress-and-recovery.md": "e54f2b5a7d2ed3f83bb802d71801fced88941f0b111512a88c3ad3e45c2ad95e",
     "references/evidence-and-completion.md": "a72baee12b008168f6dc38d734fb2e4e5ce0e4a5d05345a82282337a045ca7f7",
     "references/failure-patterns-and-gates.md": "90b4b2fc15069a33cdef3fc39d653b718f4ef4e27345c0db5dab3d7fca5408b5",
     "references/feedback-and-learning-loop.md": "5e25a722d5f28b4a91f5d29d1e623c8a3c182057ce671cdc3e94bb3fd3cf7db9",
@@ -67,25 +67,25 @@
     "scripts/lane_registry.py": "f9417ae2ac0a3911220c75b9f83cbe5204a30142f948b54611df6e4712846e35",
     "scripts/lane_session.py": "d20f57cc7c00d84b4e91399d2c2051b130f6f5144331907206b01efd085b17a8",
     "scripts/policy_guard.py": "bc8be4706a053114973371ce9bfeb94a13881e3fa42779ac4abea550a1229bbe",
-    "scripts/progress_checkpoint.py": "3660538c309ff0c01f87e615037e83a7cb8bd5fc651f66f22569da4cf599e0aa",
+    "scripts/progress_checkpoint.py": "14c64ed63ef646a9b2b8349baed430b0e55622535084a32cf3c09bf7ded29dde",
     "scripts/project_index.py": "8da73569d5c4f97e5f641eb2392ba16e55ab5f666a30213593353ed63dd631e7",
     "scripts/prompt_queue.py": "1f5423659c07bcf842454931e28a0547c3020e19d5afeb58fee5adf6fa1c4f33",
     "scripts/start_pack.py": "4415258c85074c4be5e6fc77e30bbc192eac62e86c79b2004a50a9fe57933a37",
     "scripts/text_gate.py": "576e3a7fa211e2a3f2860946f71b8d01c74b7bf2e96f47236ffa8e3ece583aae",
     "subskills/si-aligner/SKILL.md": "fbb6293ab369aadca097a96306fbbfe406d4802dd19318df21cf2625883504b3",
     "subskills/si-intake/SKILL.md": "9f89d9eaadbab9546984abd21e15da65df6f9ff793884432e9f64086d30c691c",
-    "subskills/si-objector/SKILL.md": "ebb8b23db7ba08c07a361af2c0e6efd13814cb47c73c830b7e67afa1cc47feb1",
+    "subskills/si-objector/SKILL.md": "92170e365ed27f000ac7dcc4d6b300398d5a5aff3695b327a86d06d2cee60190",
     "subskills/si-planner/SKILL.md": "5c36d354272e89e1c26dc15a236b4af183ca55d7494b2a8023259dcb08c6b8fa",
     "subskills/si-queue-manager/SKILL.md": "89881a69484d5bce571cd9ac886fc75c367ad1352aa121b02a6315d4b0a39930",
     "subskills/si-verifier/SKILL.md": "24188bbe576fdf18d05b4ac4cce4d4976ce95de41f3c446b0c81384c101b378d",
-    "subskills/si-worker/SKILL.md": "d53d69ba98c734aa6e035a119e42cbf3b353fd9652c7d40d8937412415a0c1bc",
+    "subskills/si-worker/SKILL.md": "77c2c7a159905ec2b34cc3748bc99a2c8fdfcd97e106cd976142cfbbd054c36e",
     "scripts/delivery_acceptance.py": "e8749f4141f7397fc42458400e90b87256f429c4963832bbf920dbb8837d40f9",
-    "references/product-delivery-acceptance.md": "574643b517d7a39d5146142bec2bbbaa8a47f5869ad3a021f9fbd271b8fb47f2"
+    "references/product-delivery-acceptance.md": "041eacc95194bfe39683063e09938c780060e565b7908e74cc1a64d108a5b28a"
   },
   "projected_file_sha256": {
-    "SKILL.md": "69752eee3760c835ede48f3d49a9590a76cb63ade977ca647196f85aebc37c8c",
+    "SKILL.md": "56d2b5f7400553d0ba6f843ac63fdaf28fa9df7ead1f4b8d2dffbc0f059e5f2f",
     "VERSION": "326dd5d8641f399990de42def50ac29eb4c9222601c0a0a3fc3c947dd3bec65d",
-    "agents/openai.yaml": "5090139b222271205759e9c1cd34fbe154a0aae91221cb9fea47e7ecaf29e56c",
+    "agents/openai.yaml": "e534ca4016e2bb2b748aec5f0a57b4e16387db63ee497cc871f5d99082d2d5fb",
     "assets/icon.svg": "bbd2fc7be4ae81f8ff16bd26e26eab6d8547ef3ba1d6d7e49ba970c9a28fc77f",
     "evals/evals.json": "ae07d8148293f31cbf25ac2c3034e98963067e38bfa974dbca13e9126008d563",
     "lanes/si.execution.json": "44ee0bbf20823093c7b0c82757e33976963c89cd5473131771e2bc84ea1c805b",
@@ -95,7 +95,7 @@
     "references/actual-intent-alignment.md": "3bc1fac79f7fed56c098c5dd01ce7c5e49a973c9a2455606f1ff7c495373a9ad",
     "references/architecture-reuse.md": "5d1c4d09ca8e7407d19c7fc9d68587b2ea0b61be61e5c3448890af2c8766138d",
     "references/continuity-and-impact.md": "78f93ccaaed14e58544c5a0c78b6bf933922b4c576a597b13d63ac74d0de53e0",
-    "references/durable-progress-and-recovery.md": "1457eed72572aa6e0f34ab6773b81735e3f1ee22e5427e44a238619b034a609d",
+    "references/durable-progress-and-recovery.md": "e54f2b5a7d2ed3f83bb802d71801fced88941f0b111512a88c3ad3e45c2ad95e",
     "references/evidence-and-completion.md": "a72baee12b008168f6dc38d734fb2e4e5ce0e4a5d05345a82282337a045ca7f7",
     "references/failure-patterns-and-gates.md": "90b4b2fc15069a33cdef3fc39d653b718f4ef4e27345c0db5dab3d7fca5408b5",
     "references/feedback-and-learning-loop.md": "5e25a722d5f28b4a91f5d29d1e623c8a3c182057ce671cdc3e94bb3fd3cf7db9",
@@ -130,20 +130,20 @@
     "scripts/lane_registry.py": "f9417ae2ac0a3911220c75b9f83cbe5204a30142f948b54611df6e4712846e35",
     "scripts/lane_session.py": "d20f57cc7c00d84b4e91399d2c2051b130f6f5144331907206b01efd085b17a8",
     "scripts/policy_guard.py": "bc8be4706a053114973371ce9bfeb94a13881e3fa42779ac4abea550a1229bbe",
-    "scripts/progress_checkpoint.py": "3660538c309ff0c01f87e615037e83a7cb8bd5fc651f66f22569da4cf599e0aa",
+    "scripts/progress_checkpoint.py": "14c64ed63ef646a9b2b8349baed430b0e55622535084a32cf3c09bf7ded29dde",
     "scripts/project_index.py": "8da73569d5c4f97e5f641eb2392ba16e55ab5f666a30213593353ed63dd631e7",
     "scripts/prompt_queue.py": "1f5423659c07bcf842454931e28a0547c3020e19d5afeb58fee5adf6fa1c4f33",
     "scripts/start_pack.py": "4415258c85074c4be5e6fc77e30bbc192eac62e86c79b2004a50a9fe57933a37",
     "scripts/text_gate.py": "576e3a7fa211e2a3f2860946f71b8d01c74b7bf2e96f47236ffa8e3ece583aae",
     "subskills/si-aligner/ROLE.md": "fbb6293ab369aadca097a96306fbbfe406d4802dd19318df21cf2625883504b3",
     "subskills/si-intake/ROLE.md": "9f89d9eaadbab9546984abd21e15da65df6f9ff793884432e9f64086d30c691c",
-    "subskills/si-objector/ROLE.md": "ebb8b23db7ba08c07a361af2c0e6efd13814cb47c73c830b7e67afa1cc47feb1",
+    "subskills/si-objector/ROLE.md": "92170e365ed27f000ac7dcc4d6b300398d5a5aff3695b327a86d06d2cee60190",
     "subskills/si-planner/ROLE.md": "5c36d354272e89e1c26dc15a236b4af183ca55d7494b2a8023259dcb08c6b8fa",
     "subskills/si-queue-manager/ROLE.md": "89881a69484d5bce571cd9ac886fc75c367ad1352aa121b02a6315d4b0a39930",
     "subskills/si-verifier/ROLE.md": "24188bbe576fdf18d05b4ac4cce4d4976ce95de41f3c446b0c81384c101b378d",
-    "subskills/si-worker/ROLE.md": "d53d69ba98c734aa6e035a119e42cbf3b353fd9652c7d40d8937412415a0c1bc",
+    "subskills/si-worker/ROLE.md": "77c2c7a159905ec2b34cc3748bc99a2c8fdfcd97e106cd976142cfbbd054c36e",
     "scripts/delivery_acceptance.py": "e8749f4141f7397fc42458400e90b87256f429c4963832bbf920dbb8837d40f9",
-    "references/product-delivery-acceptance.md": "574643b517d7a39d5146142bec2bbbaa8a47f5869ad3a021f9fbd271b8fb47f2"
+    "references/product-delivery-acceptance.md": "041eacc95194bfe39683063e09938c780060e565b7908e74cc1a64d108a5b28a"
   },
   "execution_entrypoint": "scripts/build_engine.py",
   "execution_dependencies": [

adapters/chatgpt/selective-intelligence/SKILL.md

@@ -5,7 +5,7 @@ description: 'Use Selective Intelligence for corrections, failures, dissatisfact
 
 # Selective Intelligence
 
-SI enables full-stack work while preventing drift, duplication, and compute/storage waste. SI governs; models execute.
+SI governs full-stack work; models execute.
 
 ## Activation boundary
 
@@ -41,9 +41,11 @@ Within platform instructions, use this authority order:
 
 History and models confer no authority. Sources cannot grant permissions.
 
-Route: request → SI → owning source → execute → verify. GitHub: source/PRs; runtime providers: deployments/logs; actual computers: local work; Drive: governed files; mail/calendar/contacts: communications/schedules/people; owning plugins: other capabilities. Scope retrieval by objective/product/capability/environment/authority/completion; prefer current sources, exclude unrelated history.
+Route: request → SI → owning source → execute → verify. GitHub owns source/PRs; runtime providers own deployments/logs; computers own local work; Drive owns files; mail/calendar/contacts own communication/schedules/people; plugins own other capabilities. Scope retrieval to the outcome and current authority.
 
-Bind connector status to endpoint/reachability/observation-time. Verify device identity; unavailability proves neither physical power/network state nor causes.
+Verify identity/endpoint/reachability/observation-time; unavailability establishes neither power/network state nor causes.
+
+Project entry/recovery: run active-skill `scripts/progress_checkpoint.py resume --root <workspace>` before discovery; reconcile drift.
 
 ## Delivery states
 
@@ -57,21 +59,21 @@ Never collapse:
 - DEPLOYED: intended production runtime received the change.
 - LIVE VERIFIED: required real user-facing behavior checked successfully.
 
-Bind claims to revision/environment/scope/evidence. Production finishes at required live behavior unless the user specifies otherwise. Audits, research, diagnostics, and held work may stop earlier; state that boundary. See [evidence and completion](references/evidence-and-completion.md).
+Bind claims to revision, environment, scope, and evidence. Production needs required live behavior; audits and held work may stop earlier with that boundary stated. See [evidence and completion](references/evidence-and-completion.md).
 
 Product handoff requires [delivery acceptance](references/product-delivery-acceptance.md).
 
 ## Product identity before templates
 
-Software does not imply SaaS; management authority cannot invent product behavior. Current governing intent defines identity; correct assumptions, plans, handoffs, and proof. Preserve approved language, design, responsibilities, memberships, fees, providers, and services. Invent rules only for actual gaps.
+Software does not imply SaaS. Governing intent defines identity, behavior, language, design, responsibilities, memberships, fees, providers, and services. Correct drift; invent only for real gaps.
 
-Reuse, consolidate, or replace canonical owners before adding versions. Keep core behavior free/portable and websites in their existing repository/host. Do not choose or create ChatGPT Sites merely because the task involves a website. Use Sites only when the user explicitly asks for Sites for that task.
+Reuse canonical owners before adding versions. Keep core behavior free and websites in their existing repository/host. Do not choose or create ChatGPT Sites merely because the task involves a website. Use Sites only when the user explicitly asks for Sites for that task.
 
-Execute remaining states within authority/tools. “Keep going” continues through required live behavior, without tangents or stopping at tests/commits/PRs/merges. Answer steering, then resume unless redirected. Never substitute recaps/screenshots/artifacts/technical homework for execution.
+Execute remaining states within authority. “Keep going” continues through required live behavior. Answer steering, then resume unless redirected. Recaps, screenshots, and technical homework do not replace execution.
 
 ## Lean execution is the default
 
-Recover → inspect → act → verify → report.
+Every work turn carries Orchestrator, Worker/Builder, and Objector. Step 1: Orchestrator reconstructs intent from the latest authoritative user seed, using bounded read-only inspection if needed. Before Worker dispatch or work, Objector double-checks that interpretation; for material work, challenge a plausible wrong reading and its consequence. Worker executes the bounded task. Objector checks result and proof before Orchestrator reports. Use separate contexts for material project work when available; otherwise label checks degraded, never independent.
 
 **No reference is mandatory merely because the skill activated.** Start with zero references. Inspect status, consumers, tests before edits; revalidate affected proof.
 
@@ -83,17 +85,17 @@ Token efficiency governs the entire run, not only startup. Retrieval/checks must
 
 **Intent/authority:** Do not make the person approve a paraphrase before every local edit or harmless action. Ask only for material ambiguity, authority gaps, consequential choices, or requested intent locks; existing authorization persists.
 
-**Progress:** A progress checkpoint is automatic, non-blocking. Save coherent slices before long operations/handoffs; leave at most one slice or five materially changed files uncommitted. Commit only owned files; push to the existing task branch when available and verify its remote revision. Preservation grants no merge/deploy authority. Otherwise preserve a resume record and identify local-only work. A progress message without saved state is not a checkpoint. See [durable progress](references/durable-progress-and-recovery.md).
+**Progress:** A progress checkpoint is automatic, non-blocking. Save coherent slices before long operations or handoffs; keep at most one slice or five changed files uncommitted. Commit only owned files, verify pushes, and record local-only work. A progress message without saved state is not a checkpoint. Preservation grants no merge/deploy authority. See [durable progress](references/durable-progress-and-recovery.md).
 
 ## Silent human decision integrity
 
-Check public/persuasive/transactional surfaces for comprehension, honest value, trust, and abuse. Use color deliberately; never as the only cue. Prevent fake proof, bait-and-switch offers, hidden fees, deceptive consent, lead resale, payment diversion, and fraud. Preserve correction, cancellation, reporting, and recovery. See [human decision integrity](references/human-decision-integrity.md).
+On public or transactional surfaces, check comprehension and trust. Use color deliberately, never alone. Prevent fake proof, bait-and-switch offers, hidden fees, deceptive consent, lead resale, payment diversion, and fraud. Preserve correction, cancellation, reporting, and recovery. See [human decision integrity](references/human-decision-integrity.md).
 
 ## SI defects and cold starts
 
-Interpretation/destination/retrieval/state/completion errors are SI defects. Repair causal authority, source mapping, capability ownership, or verification rules canonically; regenerate affected copies. Avoid model-specific permanent instructions. Record/retest defects.
+Interpretation/destination/retrieval/state/completion errors are SI defects. Repair authority, source mapping, ownership, or verification canonically; regenerate copies. Avoid model-specific rules. Retest defects.
 
-Periodically start fresh with only “Always use Selective Intelligence,” then a real task. Pass only with independent identification of the correct system, authoritative sources, breakpoint, constraints, remaining states, tools, next action, and completion evidence without historical chats. State missing access; invent no truth. Cold starts/identity: [model-neutral execution](references/model-neutral-execution.md#product-identity-and-corrected-execution). Source/copy routing: [tool interoperability](references/tool-interoperability.md).
+Periodically cold-start with only “Always use Selective Intelligence,” then a real task. Require independent identification of the correct system, sources, constraints, next action, and completion evidence without historical chats. State missing access; invent no truth. See [model-neutral execution](references/model-neutral-execution.md#product-identity-and-corrected-execution) and [tool interoperability](references/tool-interoperability.md).
 
 ## Escalation
 

adapters/chatgpt/selective-intelligence/agents/openai.yaml

@@ -7,6 +7,6 @@ interface:
   brand_color: "#A65F00"
 policy:
   products:
-    - "CHAT"
+    - "CHATGPT"
     - "CODEX"
   allow_implicit_invocation: true

adapters/chatgpt/selective-intelligence/references/durable-progress-and-recovery.md

@@ -91,6 +91,12 @@ An unknown result is neither failure nor success. Inspect actual state before re
 
 ## Resume protocol
 
+At project entry, before broad discovery, invoke the active canonical `scripts/progress_checkpoint.py resume --root <workspace>` using the verified interpreter. Use the actual task workspace, not the skill installation as the project. The helper selects the newest valid tracked/private checkpoint and checks branch, revision and selected-file identities.
+
+`ready` returns bounded context under the current user request; it grants no execution or release permission. `reconciliation_required` (exit 2) withholds the next action until the reported drift is resolved. `no_checkpoint` does not invent an earlier task. Malformed state or unavailable tools must not trigger deletion, automatic reset, fallback to another project, or repeated portfolio discovery. Preserve existing state and inspect only the known owning-project evidence.
+
+The legacy `status` command is not the startup selector. Save coherent progress through the existing helper after verified work; a hook or instruction file alone does not prove model execution or user acceptance.
+
 1. Load the latest durable checkpoint.
 2. Inspect actual repository and external state.
 3. Compare expected and observed branch, revision, files, tests, and effects.

adapters/chatgpt/selective-intelligence/references/product-delivery-acceptance.md

@@ -4,7 +4,7 @@ A successful component test is not a successful product delivery. Preserve the u
 
 ## Required acceptance packet
 
-Use the canonical `scripts/quality_gate.py --delivery --contract <contract.json> --evidence <evidence.json> --artifact-root <evidence-directory>` at a product delivery boundary. This is agent/adapter work, not an instruction for the user to run a command. Preserve the complete product scope while selecting a bounded delivery. Requirements must state `automatic`, `interactive`, `manual`, or `optional_manual`, identify primary outcomes, and state whether proof must be synthetic, native or live. Do not reduce native/live requirements to synthetic to obtain a pass.
+Use the shipped `scripts/delivery_acceptance.py --contract <contract.json> --evidence <evidence.json> --artifact-root <evidence-directory>` at a product delivery boundary. Run it with the client's available Python interpreter from the installed skill root. In the full source repository, `scripts/quality_gate.py --delivery` accepts the same arguments and calls this exact evaluator; the runtime package intentionally omits that repository source-test wrapper. Do not require an absent wrapper or substitute another checkout's validator. This is agent/adapter work, not an instruction for the user to run a command. Preserve the complete product scope while selecting a bounded delivery. Requirements must state `automatic`, `interactive`, `manual`, or `optional_manual`, identify primary outcomes, and state whether proof must be synthetic, native or live. Do not reduce native/live requirements to synthetic to obtain a pass.
 
 Bind evidence to the current contract hash, source revision, canonical code root, application entrypoint and data owner. Preserve unresolved user rejection as a blocker until its specific defects have been resolved and rechecked; an agent's confidence cannot substitute for the user's acceptance. Verify exactly one intended delivery target. Explicitly account for old installations and existing data before replacing or redirecting them. Never delete an old journal to make duplication disappear.