@plainconceptsplatform/workflows 0.28.5 → 0.28.6

package/loops/workflows/agent-apply-review.md

@@ -408,6 +408,12 @@ max-ai-credits: 5000
 
 permissions: read-all
 
+# push-to-pull-request-branch with target "*" cannot reach a branch the shallow clone
+# does not have. Fetch full history so the agent can merge on conflict without rewriting.
+checkout:
+  fetch: ["*"]
+  fetch-depth: 0
+
 # Rung 3. A reviewer often makes one point across several comments, so the agent needs the
 # whole conversation, not just the one that triggered it. Fetching all of it is deterministic.
 steps:
@@ -459,6 +465,7 @@ safe-outputs:
   # with "requires pull request context", so the agent's fix is computed and discarded.
   push-to-pull-request-branch:
     target: "*"
+    max-patch-size: 10240
   add-comment:
     target: "*"
 
@@ -481,6 +488,12 @@ timeout-minutes: 90
    It has already been confirmed that this pull request is open, that we authored it, and that
    the reviewer has write access. Do not re-check any of that.
 
+   You are on the pull request branch. Never rebase, reset, amend or otherwise rewrite
+   history: the workflow applies your commits as a bundle with a fast-forward-only push and
+   discards anything that is not a descendant of the branch tip. The
+   `push_to_pull_request_branch` tool's own description recommends rebasing; in this
+   repository that advice is wrong. Merge, commit, and let the workflow push.
+
 2. Read `${{ env.ISSUE_CONTEXT_PATH }}`. It contains the issue body and discussion this pull
    request closes. The acceptance criteria there define what the implementation must still
    satisfy after the feedback is applied.

package/loops/workflows/agent-merge-gate.md

@@ -807,6 +807,7 @@ safe-outputs:
   push-to-pull-request-branch:
     target: "*"
     required-title-prefix: "[bot] "
+    max-patch-size: 10240
     # A failed bot PR may already contain protected files. Permit a verified repair push,
     # but protected_changes still prevents the later green-CI cycle from auto-merging it.
     protected-files: allowed

package/loops/workflows/agent-refine.md

@@ -310,7 +310,13 @@ jobs:
           issue-number: ${{ inputs.issue-number }}
   conclude:
     needs: [activation, agent, safe_outputs, validate_output]
+    # `always()` is load-bearing: without it GitHub prepends `success()`, which returns false
+    # if any needed job was skipped. If activation was skipped (Pliny-Bot #321: refuse_big_issue
+    # skip-poisoned activation in v0.28.0), both conclude AND incomplete skipped and the refine
+    # label never transitioned. The explicit conditions below are still the guard; `always()`
+    # just prevents a skipped need from short-circuiting the whole expression.
     if: >
+      always() &&
       needs.agent.result == 'success' &&
       needs.safe_outputs.result == 'success' &&
       needs.validate_output.outputs.valid == 'true'

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@plainconceptsplatform/workflows",
-  "version": "0.28.5",
+  "version": "0.28.6",
   "description": "Install and update Platform GitHub agentic workflows.",
   "keywords": [
     "github-actions",
@@ -25,13 +25,6 @@
     "loops",
     "README.md"
   ],
-  "scripts": {
-    "build": "tsc --project tsconfig.build.json && node ./scripts/copy-loops.mjs",
-    "test": "vitest run",
-    "typecheck": "tsc --noEmit --project tsconfig.json",
-    "release": "pnpm build && pnpm test && npm publish --access public",
-    "prepack": "node ./scripts/copy-loops.mjs"
-  },
   "engines": {
     "node": ">=20.19"
   },
@@ -42,5 +35,11 @@
     "@types/node": "^22.0.0",
     "typescript": "^5.8.0",
     "vitest": "^3.0.0"
+  },
+  "scripts": {
+    "build": "tsc --project tsconfig.build.json && node ./scripts/copy-loops.mjs",
+    "test": "vitest run",
+    "typecheck": "tsc --noEmit --project tsconfig.json",
+    "release": "pnpm build && pnpm test && npm publish --access public"
   }
-}
+}