@harness-forge/cli 1.1.1 → 1.2.0

package/docs/agents.md

@@ -14,6 +14,9 @@ AI content into the hidden `.hforge/` layer.
 - `.hforge/library/rules/` and `.hforge/library/knowledge/` for hidden installed rule and knowledge surfaces
 - `.hforge/runtime/recursive/sessions/` for optional hard-task recursive
   sessions that stay hidden alongside the rest of the AI layer
+- `.hforge/runtime/recursive/language-capabilities.json` for the canonical
+  recursive structured-analysis capability map before deeper recursive
+  investigation begins
 - `docs/authoring/enhanced-skill-import.md` for curated research and validation provenance behind imported skill upgrades
 - `RESEARCH-SOURCES.md` and `VALIDATION.md` for optional pack-level provenance detail
 
@@ -29,6 +32,8 @@ AI content into the hidden `.hforge/` layer.
 - when the application repo should stay clean and treat AI-only content as a hidden support layer
 - when a task is multi-hop, ambiguous, or architecture-heavy enough to justify
   an explicit recursive session instead of growing the prompt
+- when a recursive session needs one bounded structured analysis step with a
+  durable run record instead of chat-only reasoning
 
 ## Discovery rule
 
@@ -39,6 +44,14 @@ AI content into the hidden `.hforge/` layer.
 - in installed workspaces, recursive runtime state lives under
   `.hforge/runtime/recursive/sessions/<sessionId>/` and should be treated as
   hidden operational state rather than product code
+- use recursive structured analysis as the promoted execution surface for
+  recursive investigation, and treat the REPL as optional rather than
+  authoritative
+- treat `.hforge/runtime/recursive/language-capabilities.json` as the honest
+  recursive support contract before claiming a language has deep or native
+  structured-analysis support
+- treat Cursor and OpenCode recursive promotion as translated shared-runtime
+  guidance, not native parity with Codex or Claude Code
 - for custom agents that do not want to parse prose only, load
   `.hforge/agent-manifest.json` first and treat every surface it marks
   `treatAsProductCode: false` as AI-layer content rather than application code

package/docs/commands.md

@@ -23,6 +23,7 @@ maintenance commands through the `hforge` CLI.
 - `shell setup` and `shell status` for optional PATH integration without forcing a global npm install
 - `status` for current workspace state
 - `refresh` for shared runtime regeneration after install or maintenance work
+- `update` and `upgrade` for non-destructive package refresh from the latest published Harness Forge version while preserving gathered runtime state
 - `task` for task-runtime inspection
 - `pack` for task-pack inspection
 - `review` for runtime and decision-health summaries
@@ -36,9 +37,11 @@ maintenance commands through the `hforge` CLI.
   Cursor, and OpenCode support inspection
 - `template validate` and `template suggest` for template and workflow flows
 - `flow status` for recoverable Speckit state
-- `recursive plan`, `recursive inspect`, `recursive adr`, `recursive resume`,
-  `recursive finalize`, `recursive compact`, and `recursive repl` for optional
-  recursive-runtime operation on difficult work
+- `recursive plan`, `recursive capabilities`, `recursive run`,
+  `recursive runs`, `recursive inspect-run`, `recursive inspect`,
+  `recursive adr`, `recursive resume`, `recursive finalize`,
+  `recursive compact`, and `recursive repl` for optional recursive-runtime
+  operation on difficult work
 - `observability summarize` and `observability report` for local metrics and
   effectiveness review
 - `parallel plan`, `parallel status`, and `parallel merge-check` for shard
@@ -53,6 +56,12 @@ maintenance commands through the `hforge` CLI.
 Use bare `hforge` after `shell setup`, after a global install, or by replacing
 it with the workspace-local launcher under `.hforge/generated/bin/`.
 
+For agents inside an installed workspace, the safest execution order is:
+
+1. `.hforge/generated/bin/hforge.cmd` or `.ps1` on Windows, or `./.hforge/generated/bin/hforge` on POSIX
+2. bare `hforge`
+3. `npx @harness-forge/cli`
+
 ```bash
 npx @harness-forge/cli
 npx @harness-forge/cli shell setup --yes
@@ -63,6 +72,8 @@ hforge shell status --json
 hforge init --root . --json
 hforge bootstrap --root . --yes
 hforge refresh --root . --json
+hforge update --root . --dry-run --yes
+hforge update --root . --yes
 hforge task list --root . --json
 hforge task inspect TASK-001 --root . --json
 hforge pack inspect TASK-001 --root . --json
@@ -81,6 +92,10 @@ hforge template validate --json
 hforge template suggest bugfix
 hforge flow status --json
 hforge recursive plan "investigate billing retry behavior" --task-id TASK-001 --json
+hforge recursive capabilities --root . --json
+hforge recursive run RS-123 --file analyze-billing.mjs --json
+hforge recursive runs RS-123 --json
+hforge recursive inspect-run RS-123 RUN-001 --json
 hforge recursive inspect RS-123 --json
 hforge observability summarize --json
 hforge observability report . --json
@@ -95,6 +110,12 @@ hforge upgrade-surface --json
 hforge prune --json
 ```
 
+The `update` and `upgrade` commands download the requested published package
+version or dist-tag, reapply managed surfaces, write an install-state backup
+under `.hforge/state/`, and intentionally preserve gathered runtime state such
+as task artifacts, decision records, recursive sessions, and observability
+signals.
+
 ## Maintainer source-checkout examples
 
 When developing Harness Forge itself from a source checkout, the equivalent
@@ -177,6 +198,10 @@ hforge parallel merge-check --json
 
 ```bash
 hforge recursive plan "investigate billing retry behavior across the route and service" --task-id TASK-001 --json
+hforge recursive capabilities --root . --json
+hforge recursive run RS-123 --file analyze-billing.mjs --json
+hforge recursive runs RS-123 --json
+hforge recursive inspect-run RS-123 RUN-001 --json
 hforge recursive inspect RS-123 --json
 hforge recursive compact RS-123 --json
 hforge recursive finalize RS-123 --json
@@ -187,6 +212,11 @@ The `recursive plan` entrypoint writes a durable draft session under
 budget policy, the seeded handles, and the current promotion state, and leaves
 ordinary non-recursive task/runtime flows untouched.
 
+The promoted recursive structured-analysis path uses
+`.hforge/runtime/recursive/language-capabilities.json` as the canonical
+workspace capability map and records run artifacts under
+`.hforge/runtime/recursive/sessions/<sessionId>/runs/`.
+
 ### Inspect the hidden runtime after initialization
 
 ```bash

package/docs/generated-artifacts.md

@@ -85,8 +85,19 @@ explicit and recoverable.
   template registry for task-time artifacts
 - `.hforge/runtime/recursive/sessions/RS-XXX/session.json` is the durable draft
   or active recursive session identity, budget, handles, and tool surface
+- `.hforge/runtime/recursive/language-capabilities.json` is the canonical
+  recursive structured-analysis capability map for language adapter depth,
+  degraded states, and native-execution posture
+- `.hforge/runtime/recursive/sessions/RS-XXX/execution-policy.json` is the
+  session-scoped structured-execution posture for recursive analysis runs
+- `.hforge/runtime/recursive/sessions/RS-XXX/capabilities.json` is the
+  session-visible recursive capability snapshot used by structured runs
 - `.hforge/runtime/recursive/sessions/RS-XXX/memory.json` is compact
   recursive-session working memory for the current investigation
+- `.hforge/runtime/recursive/sessions/RS-XXX/runs/RUN-XXX/meta.json` is the
+  durable metadata record for one structured recursive analysis run
+- `.hforge/runtime/recursive/sessions/RS-XXX/runs/RUN-XXX/result.json` is the
+  machine-readable result for one structured recursive analysis run
 - `.hforge/runtime/recursive/sessions/RS-XXX/trace.jsonl` is append-only
   recursive trace output for operator auditability
 - `.hforge/runtime/recursive/sessions/RS-XXX/summary.json` is the deterministic
@@ -94,6 +105,8 @@ explicit and recoverable.
 - `.specify/state/flow-state.json` is runtime state, not an authored source
 - `.hforge/state/install-state.json` is generated install-state and runtime-version metadata, not authored product content
 - `.hforge/state/post-install-guidance.txt` is generated operator guidance for init, install, and recovery flows
+- `.hforge/state/update-backup-*.json` is the pre-update install-state snapshot
+  written before a non-destructive `hforge update` or `hforge upgrade` apply run
 
 ## Artifact lineage rules
 

package/docs/install/targets.md

@@ -17,6 +17,7 @@ The runtime now hydrates baseline repo-intelligence artifacts during bootstrap:
 - `.hforge/runtime/repo/recommendations.json`
 - `.hforge/runtime/repo/instruction-plan.json`
 - `.hforge/runtime/findings/risk-signals.json`
+- `.hforge/runtime/recursive/language-capabilities.json`
 
 Canonical target support truth lives in
 `manifests/catalog/harness-capability-matrix.json`.
@@ -27,3 +28,5 @@ Use it when you need to answer:
 - what fallback behavior degraded support requires
 - whether a support claim is still aligned with the generated
   `docs/target-support-matrix.md` surface
+- whether recursive structured analysis is promoted natively or only through
+  translated shared-runtime guidance for the selected target

package/docs/installation.md

@@ -47,8 +47,15 @@ After install changes, use:
 
 ```bash
 hforge refresh --root /path/to/your/workspace --json
+hforge update --root /path/to/your/workspace --dry-run --yes
+hforge update --root /path/to/your/workspace --yes
 ```
 
+The non-destructive updater downloads the requested published Harness Forge
+version or dist-tag, reapplies managed surfaces, writes an install-state backup
+under `.hforge/state/`, and preserves gathered runtime state such as task
+artifacts, decision records, recursive sessions, and observability outputs.
+
 Every successful install also writes a shared runtime summary under
 `.hforge/runtime/` and materializes the canonical hidden AI layer under
 `.hforge/library/` and `.hforge/templates/`:
@@ -76,6 +83,8 @@ Every successful install also writes a shared runtime summary under
 - `.hforge/runtime/recursive/sessions/RS-XXX/session.json` appears only when an
   operator explicitly escalates hard work into recursive mode and records the
   durable draft session, budget, handles, and promotion state
+- `.hforge/runtime/recursive/language-capabilities.json` records the canonical
+  recursive structured-analysis capability truth for the workspace
 
 ## Shell integration vs global install
 

package/docs/maintenance-lifecycle.md

@@ -14,6 +14,7 @@ Harness Forge maintenance should answer three questions quickly:
 | `audit` | detailed install-state summary |
 | `diff-install` | compare recorded managed files against what still exists |
 | `refresh` | rewrite shared runtime summaries and baseline repo-intelligence artifacts |
+| `update` / `upgrade` | download the requested published Harness Forge version and reapply managed surfaces while preserving gathered runtime state |
 | `sync` | normalize duplicated install-state entries |
 | `upgrade-surface` | explain how to refresh an older install safely |
 | `prune` | identify duplicate recorded writes before cleanup |
@@ -24,9 +25,10 @@ Harness Forge maintenance should answer three questions quickly:
 1. run `doctor` to spot obvious drift
 2. run `audit` for the full install-state summary
 3. use `refresh` when installed targets are correct but shared runtime summaries drifted
-4. use `diff-install` and `sync` to understand state mismatch
-4. use `upgrade-surface` before refreshing an older install
-5. use `prune` only after reviewing the candidates
+4. use `update --dry-run --yes` to preview a non-destructive package refresh when you want newer managed content
+5. use `diff-install` and `sync` to understand state mismatch
+6. use `upgrade-surface` before refreshing an older install if you only need guidance
+7. use `prune` only after reviewing the candidates
 
 ## Example commands
 
@@ -34,6 +36,8 @@ Harness Forge maintenance should answer three questions quickly:
 hforge doctor --json
 hforge audit --json
 hforge refresh --root . --json
+hforge update --root . --dry-run --yes
+hforge update --root . --yes
 hforge diff-install --json
 hforge sync --json
 hforge upgrade-surface --json
@@ -44,6 +48,7 @@ hforge prune --json
 
 - maintenance commands should be diagnostic first and destructive second
 - drift should point to concrete files or bundle ids
+- update should preserve gathered runtime state such as task artifacts, decision indexes, recursive sessions, and observability signals
 - upgrade guidance should tell the operator what to rerun next
 - prune should never remove state silently without explicit opt-in
 - refresh should be safe to rerun and should not require reinstalling bundles just to regenerate runtime summaries

package/docs/quickstart.md

@@ -8,6 +8,8 @@ npx @harness-forge/cli init --root . --agent codex --dry-run
 npx @harness-forge/cli bootstrap --root .
 npx @harness-forge/cli init --root . --json
 hforge commands --json
+hforge recursive capabilities --root . --json
+hforge update --root . --dry-run --yes
 npm run validate:release
 ```
 
@@ -37,5 +39,6 @@ npm run validate:release
 6. choose a target (`claude-code` or `codex`)
 7. preview the install plan for the language and capability bundles you need
 8. inspect an existing repository with `recommend` when the right pack is not obvious
-9. use `refresh`, `task`, `pack`, `review`, and `export` to inspect the hidden runtime once work is in flight
-10. use `.specify/` and the promoted skills for real work
+9. use `refresh`, `task`, `pack`, `review`, `export`, and `recursive capabilities` to inspect the hidden runtime once work is in flight
+10. use `update --dry-run --yes` before refreshing to the latest published Harness Forge package in place
+11. use `.specify/` and the promoted skills for real work

package/docs/release-process.md

@@ -16,6 +16,8 @@
 - `npm run validate:doc-command-alignment`
 - `npm run validate:runtime-consistency`
 - `npm run validate:release`
+- confirm recursive structured-analysis docs and generated-artifact surfaces stay aligned
+- confirm `hforge update --root <repo> --dry-run --yes` still previews a safe non-destructive refresh path
 
 ## Publish posture
 

package/docs/target-support-matrix.md

@@ -24,10 +24,10 @@ cross-surface relationships.
 
 | Target | Overall support | Strengths | Gaps |
 | --- | --- | --- | --- |
-| `codex` | full | Templates and docs, Workflow validation, Repo intelligence, Flow orchestration, Maintenance commands, Local observability | Typed hooks |
-| `claude-code` | full | Templates and docs, Workflow validation, Repo intelligence, Flow orchestration, Typed hooks, Maintenance commands, Local observability | none |
-| `cursor` | partial | none | Templates and docs, Workflow validation, Repo intelligence, Flow orchestration, Typed hooks, Maintenance commands, Local observability |
-| `opencode` | partial | none | Templates and docs, Workflow validation, Repo intelligence, Flow orchestration, Typed hooks, Maintenance commands, Local observability |
+| `codex` | full | Templates and docs, Workflow validation, Repo intelligence, Recursive structured analysis, Flow orchestration, Maintenance commands, Local observability | Typed hooks |
+| `claude-code` | full | Templates and docs, Workflow validation, Repo intelligence, Recursive structured analysis, Flow orchestration, Typed hooks, Maintenance commands, Local observability | none |
+| `cursor` | partial | none | Templates and docs, Workflow validation, Repo intelligence, Recursive structured analysis, Flow orchestration, Typed hooks, Maintenance commands, Local observability |
+| `opencode` | partial | none | Templates and docs, Workflow validation, Repo intelligence, Recursive structured analysis, Flow orchestration, Typed hooks, Maintenance commands, Local observability |
 
 ## Capability honesty
 
@@ -36,6 +36,7 @@ cross-surface relationships.
 | Templates and docs | full (native) | full (native) | partial (documentation-only) | partial (documentation-only) |
 | Workflow validation | full (native) | full (native) | partial (documentation-only) | partial (documentation-only) |
 | Repo intelligence | full (native) | full (native) | partial (translated) | partial (translated) |
+| Recursive structured analysis | full (native) | full (native) | partial (translated) | partial (translated) |
 | Flow orchestration | full (native) | full (native) | unsupported (unsupported) | unsupported (unsupported) |
 | Typed hooks | partial (documentation-only) | full (native) | unsupported (unsupported) | unsupported (unsupported) |
 | Maintenance commands | full (native) | full (native) | partial (documentation-only) | partial (documentation-only) |
@@ -56,6 +57,7 @@ cross-surface relationships.
 - Templates and docs: partial via documentation-only. Fallback: Use generated docs and manifests as the support surface instead of expecting a native Cursor runtime install. Cursor can consume packaged docs and portable guidance, but not a first-class runtime mapping.
 - Workflow validation: partial via documentation-only. Fallback: Run the validators from the repository rather than relying on a Cursor-specific runtime integration. Validation guidance is portable, but no Cursor-native runtime wiring is shipped.
 - Repo intelligence: partial via translated. Fallback: Inspect `.hforge/runtime/index.json` and `.hforge/library/`, then run the intelligence scripts directly and treat the output as operator guidance. Repo intelligence is available as thin bridges, portable scripts, and generated hidden `.hforge/` state rather than a native Cursor runtime surface.
+- Recursive structured analysis: partial via translated. Fallback: Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native Cursor parity. Cursor receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload.
 - Flow orchestration: unsupported via unsupported. Fallback: Do not rely on packaged flow recovery for Cursor; use `.hforge/runtime/index.json` and the thin bridge surfaces only. No first-class Cursor flow-state recovery surface is shipped beyond the hidden `.hforge/` runtime summary and bridge surfaces.
 - Typed hooks: unsupported via unsupported. Fallback: Use hook docs as references only; do not claim native Cursor hook execution. Cursor does not have a packaged native hook runtime in this repository.
 - Maintenance commands: partial via documentation-only. Fallback: Run the maintenance commands from the repository rather than expecting Cursor-specific runtime integration. Maintenance guidance is portable, but a Cursor-native runtime surface is not shipped.
@@ -66,6 +68,7 @@ cross-surface relationships.
 - Templates and docs: partial via documentation-only. Fallback: Use generated docs and manifests as the support surface instead of expecting a native OpenCode runtime install. OpenCode can consume packaged docs and portable guidance, but not a first-class runtime mapping.
 - Workflow validation: partial via documentation-only. Fallback: Run the validators from the repository rather than relying on an OpenCode-specific runtime integration. Validation guidance is portable, but no OpenCode-native runtime wiring is shipped.
 - Repo intelligence: partial via translated. Fallback: Inspect `.hforge/runtime/index.json` and `.hforge/library/`, then run the intelligence scripts directly and treat the output as operator guidance. Repo intelligence is available as thin bridges, portable scripts, and generated hidden `.hforge/` state rather than a native OpenCode runtime surface.
+- Recursive structured analysis: partial via translated. Fallback: Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native OpenCode parity. OpenCode receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload.
 - Flow orchestration: unsupported via unsupported. Fallback: Do not rely on packaged flow recovery for OpenCode; use `.hforge/runtime/index.json` and the thin bridge surfaces only. No first-class OpenCode flow-state recovery surface is shipped beyond the hidden `.hforge/` runtime summary and bridge surfaces.
 - Typed hooks: unsupported via unsupported. Fallback: Use hook docs as references only; do not claim native OpenCode hook execution. OpenCode does not have a packaged native hook runtime in this repository.
 - Maintenance commands: partial via documentation-only. Fallback: Run the maintenance commands from the repository rather than expecting OpenCode-specific runtime integration. Maintenance guidance is portable, but an OpenCode-native runtime surface is not shipped.

package/manifests/catalog/capability-taxonomy.json

@@ -19,6 +19,12 @@
       "displayName": "Repo intelligence",
       "description": "Repository scanning, framework detection, and evidence-backed recommendation output."
     },
+    {
+      "id": "recursive-structured-analysis",
+      "family": "orchestration",
+      "displayName": "Recursive structured analysis",
+      "description": "Session-scoped structured recursive analysis, capability truth, durable run records, and related bridge guidance."
+    },
     {
       "id": "flow-orchestration",
       "family": "orchestration",

package/manifests/catalog/compatibility-matrix.json

@@ -1,5 +1,5 @@
 {
-  "generatedAt": "2026-03-28T16:55:31.339Z",
+  "generatedAt": "2026-03-28T21:42:25.027Z",
   "entries": [
     {
       "subjectType": "language",
@@ -775,6 +775,23 @@
       "fallbackBehavior": "Inspect .hforge/observability artifacts directly.",
       "notes": "Local observability is available for Claude Code workflows."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "claude-code",
+      "relationType": "supports-capability",
+      "relatedType": "capability",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "full",
+      "supportMode": "native",
+      "confidence": 0.96,
+      "evidenceSource": [
+        "src/cli/commands/recursive.ts",
+        "src/application/recursive/run-structured-analysis.ts",
+        "targets/claude-code/adapter.json"
+      ],
+      "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json`, recursive session artifacts, and the promoted recursive CLI surfaces.",
+      "notes": "Claude Code promotes recursive structured analysis through the shared runtime capability map, structured run commands, and durable session artifacts."
+    },
     {
       "subjectType": "target",
       "subjectId": "claude-code",
@@ -1055,6 +1072,14 @@
       "relatedId": "python-engineering",
       "supportLevel": "full"
     },
+    {
+      "subjectType": "target",
+      "subjectId": "claude-code",
+      "relationType": "supports",
+      "relatedType": "skill",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "full"
+    },
     {
       "subjectType": "target",
       "subjectId": "claude-code",
@@ -1545,6 +1570,23 @@
       "fallbackBehavior": "Inspect .hforge/observability artifacts directly.",
       "notes": "Local observability is available for Codex workflows."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "codex",
+      "relationType": "supports-capability",
+      "relatedType": "capability",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "full",
+      "supportMode": "native",
+      "confidence": 0.95,
+      "evidenceSource": [
+        "src/cli/commands/recursive.ts",
+        "src/application/recursive/run-structured-analysis.ts",
+        "targets/codex/adapter.json"
+      ],
+      "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json`, recursive session artifacts, and the promoted recursive CLI surfaces.",
+      "notes": "Codex promotes recursive structured analysis through the shared runtime capability map, structured run commands, and durable session artifacts."
+    },
     {
       "subjectType": "target",
       "subjectId": "codex",
@@ -1825,6 +1867,14 @@
       "relatedId": "python-engineering",
       "supportLevel": "full"
     },
+    {
+      "subjectType": "target",
+      "subjectId": "codex",
+      "relationType": "supports",
+      "relatedType": "skill",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "full"
+    },
     {
       "subjectType": "target",
       "subjectId": "codex",
@@ -2349,6 +2399,23 @@
       "fallbackBehavior": "Inspect the local observability artifacts directly from the repository.",
       "notes": "Observability artifacts are portable, but no native Cursor runtime surface is shipped."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "cursor",
+      "relationType": "supports-capability",
+      "relatedType": "capability",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "partial",
+      "supportMode": "translated",
+      "confidence": 0.84,
+      "evidenceSource": [
+        "targets/cursor/adapter.json",
+        "src/cli/commands/recursive.ts",
+        "docs/agents.md"
+      ],
+      "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native Cursor parity.",
+      "notes": "Cursor receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload."
+    },
     {
       "subjectType": "target",
       "subjectId": "cursor",
@@ -2656,6 +2723,15 @@
       "supportLevel": "partial",
       "notes": "Cursor only provides partial skill support for python-engineering."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "cursor",
+      "relationType": "supports",
+      "relatedType": "skill",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "partial",
+      "notes": "Cursor only provides partial skill support for recursive-structured-analysis."
+    },
     {
       "subjectType": "target",
       "subjectId": "cursor",
@@ -3199,6 +3275,23 @@
       "fallbackBehavior": "Inspect the local observability artifacts directly from the repository.",
       "notes": "Observability artifacts are portable, but no native OpenCode runtime surface is shipped."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "opencode",
+      "relationType": "supports-capability",
+      "relatedType": "capability",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "partial",
+      "supportMode": "translated",
+      "confidence": 0.84,
+      "evidenceSource": [
+        "targets/opencode/adapter.json",
+        "src/cli/commands/recursive.ts",
+        "docs/agents.md"
+      ],
+      "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native OpenCode parity.",
+      "notes": "OpenCode receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload."
+    },
     {
       "subjectType": "target",
       "subjectId": "opencode",
@@ -3506,6 +3599,15 @@
       "supportLevel": "partial",
       "notes": "OpenCode only provides partial skill support for python-engineering."
     },
+    {
+      "subjectType": "target",
+      "subjectId": "opencode",
+      "relationType": "supports",
+      "relatedType": "skill",
+      "relatedId": "recursive-structured-analysis",
+      "supportLevel": "partial",
+      "notes": "OpenCode only provides partial skill support for recursive-structured-analysis."
+    },
     {
       "subjectType": "target",
       "subjectId": "opencode",

package/manifests/catalog/flow-artifacts.json

@@ -288,6 +288,30 @@
       "required": false,
       "lineageRole": "recursive-session"
     },
+    {
+      "id": "recursive-runtime-language-capabilities",
+      "stage": "shared",
+      "pathPattern": ".hforge/runtime/recursive/language-capabilities.json",
+      "generatedBy": "recursive-runtime",
+      "required": false,
+      "lineageRole": "recursive-language-capabilities"
+    },
+    {
+      "id": "recursive-runtime-execution-policy",
+      "stage": "shared",
+      "pathPattern": ".hforge/runtime/recursive/sessions/*/execution-policy.json",
+      "generatedBy": "recursive-runtime",
+      "required": false,
+      "lineageRole": "recursive-execution-policy"
+    },
+    {
+      "id": "recursive-runtime-session-capabilities",
+      "stage": "shared",
+      "pathPattern": ".hforge/runtime/recursive/sessions/*/capabilities.json",
+      "generatedBy": "recursive-runtime",
+      "required": false,
+      "lineageRole": "recursive-session-capabilities"
+    },
     {
       "id": "recursive-runtime-memory",
       "stage": "shared",
@@ -312,6 +336,22 @@
       "required": false,
       "lineageRole": "recursive-summary"
     },
+    {
+      "id": "recursive-runtime-run-meta",
+      "stage": "shared",
+      "pathPattern": ".hforge/runtime/recursive/sessions/*/runs/*/meta.json",
+      "generatedBy": "recursive-runtime",
+      "required": false,
+      "lineageRole": "recursive-structured-run"
+    },
+    {
+      "id": "recursive-runtime-run-result",
+      "stage": "shared",
+      "pathPattern": ".hforge/runtime/recursive/sessions/*/runs/*/result.json",
+      "generatedBy": "recursive-runtime",
+      "required": false,
+      "lineageRole": "recursive-structured-run-result"
+    },
     {
       "id": "flow-state",
       "stage": "shared",

package/manifests/catalog/harness-capability-matrix.json

@@ -1,7 +1,7 @@
 {
   "$schema": "../../schemas/manifests/harness-capability-matrix.schema.json",
   "version": 1,
-  "generatedAt": "2026-03-25T00:00:00.000Z",
+  "generatedAt": "2026-03-28T00:00:00.000Z",
   "targets": [
     {
       "targetId": "codex",
@@ -50,6 +50,21 @@
           "notes": "Repo intelligence is a first-class Codex surface and compiles into the hidden `.hforge/` AI layer.",
           "fallbackBehavior": "Inspect `.hforge/runtime/index.json` and `.hforge/library/`, or run the intelligence scripts directly if the CLI is unavailable."
         },
+        {
+          "capabilityId": "recursive-structured-analysis",
+          "supportLevel": "full",
+          "supportMode": "native",
+          "evidenceSource": [
+            "src/cli/commands/recursive.ts",
+            "src/application/recursive/run-structured-analysis.ts",
+            "targets/codex/adapter.json"
+          ],
+          "lastValidatedAt": "2026-03-28T00:00:00.000Z",
+          "validationMethod": "recursive-runtime-and-adapter-review",
+          "confidence": 0.95,
+          "notes": "Codex promotes recursive structured analysis through the shared runtime capability map, structured run commands, and durable session artifacts.",
+          "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json`, recursive session artifacts, and the promoted recursive CLI surfaces."
+        },
         {
           "capabilityId": "flow-orchestration",
           "supportLevel": "full",
@@ -156,6 +171,21 @@
           "notes": "Repo intelligence is a first-class Claude Code surface and compiles into the hidden `.hforge/` AI layer.",
           "fallbackBehavior": "Inspect `.hforge/runtime/index.json` and `.hforge/library/`, or run the intelligence scripts directly if the CLI is unavailable."
         },
+        {
+          "capabilityId": "recursive-structured-analysis",
+          "supportLevel": "full",
+          "supportMode": "native",
+          "evidenceSource": [
+            "src/cli/commands/recursive.ts",
+            "src/application/recursive/run-structured-analysis.ts",
+            "targets/claude-code/adapter.json"
+          ],
+          "lastValidatedAt": "2026-03-28T00:00:00.000Z",
+          "validationMethod": "recursive-runtime-and-adapter-review",
+          "confidence": 0.96,
+          "notes": "Claude Code promotes recursive structured analysis through the shared runtime capability map, structured run commands, and durable session artifacts.",
+          "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json`, recursive session artifacts, and the promoted recursive CLI surfaces."
+        },
         {
           "capabilityId": "flow-orchestration",
           "supportLevel": "full",
@@ -262,6 +292,21 @@
           "notes": "Repo intelligence is available as thin bridges, portable scripts, and generated hidden `.hforge/` state rather than a native Cursor runtime surface.",
           "fallbackBehavior": "Inspect `.hforge/runtime/index.json` and `.hforge/library/`, then run the intelligence scripts directly and treat the output as operator guidance."
         },
+        {
+          "capabilityId": "recursive-structured-analysis",
+          "supportLevel": "partial",
+          "supportMode": "translated",
+          "evidenceSource": [
+            "targets/cursor/adapter.json",
+            "src/cli/commands/recursive.ts",
+            "docs/agents.md"
+          ],
+          "lastValidatedAt": "2026-03-28T00:00:00.000Z",
+          "validationMethod": "shared-runtime-promotion-review",
+          "confidence": 0.84,
+          "notes": "Cursor receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload.",
+          "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native Cursor parity."
+        },
         {
           "capabilityId": "flow-orchestration",
           "supportLevel": "unsupported",
@@ -367,6 +412,21 @@
           "notes": "Repo intelligence is available as thin bridges, portable scripts, and generated hidden `.hforge/` state rather than a native OpenCode runtime surface.",
           "fallbackBehavior": "Inspect `.hforge/runtime/index.json` and `.hforge/library/`, then run the intelligence scripts directly and treat the output as operator guidance."
         },
+        {
+          "capabilityId": "recursive-structured-analysis",
+          "supportLevel": "partial",
+          "supportMode": "translated",
+          "evidenceSource": [
+            "targets/opencode/adapter.json",
+            "src/cli/commands/recursive.ts",
+            "docs/agents.md"
+          ],
+          "lastValidatedAt": "2026-03-28T00:00:00.000Z",
+          "validationMethod": "shared-runtime-promotion-review",
+          "confidence": 0.84,
+          "notes": "OpenCode receives recursive structured-analysis promotion through shared-runtime capability maps and durable session artifacts rather than a native target runtime payload.",
+          "fallbackBehavior": "Use `.hforge/runtime/recursive/language-capabilities.json` and recursive session artifacts as the honest support surface instead of claiming native OpenCode parity."
+        },
         {
           "capabilityId": "flow-orchestration",
           "supportLevel": "unsupported",