PromptBP Task Environment Compiler

Recorded connector-delivered creator test — October 22, 2026

Planning completed with ALLOW for the scoped review. The planned task remains not_started and settings remain applied:false. Below is the complete generated plan; only the machine-local checkout path is replaced with OWNER_WORKSPACE/PromptBP. This displayed redacted copy has different bytes from the original artifact bound by the retained local planning receipt. No tool configuration, agent dispatch, purchase or task execution occurred as a planning side effect.

{
  "schema_version": "promptbp-agent-0.3",
  "plan_id": "promptbp-live-use-2026-10-22-01",
  "objective": "Review the current three-file PromptBP scope-fix patch read-only and verify its focused regression checks.",
  "task_class": "repository_review",
  "authority": {
    "mode": "read_only_repository; permitted_local_evidence_output",
    "ref": "Active user goal and approved agent-native delivery; scoped validation remains authorized."
  },
  "scope": [
    "OWNER_WORKSPACE/PromptBP"
  ],
  "route": {
    "host": "Codex desktop",
    "source": "workspace",
    "available": true
  },
  "environment": {
    "model": {
      "selection": "active_session",
      "id": "unknown",
      "thinking": "unknown",
      "basis": "Exact runtime settings not exposed; no fixed model requested and no setting changed."
    },
    "selected": [
      {
        "kind": "tool",
        "id": "functions.exec/exec_command",
        "roles": [
          "repository_read",
          "file_read",
          "bounded_test_execution",
          "evidence_hash"
        ],
        "basis": "Available native PowerShell/Git/Python host tools; focused test effects inspected."
      },
      {
        "kind": "context",
        "id": "current-user-AGENTS-and-workspace",
        "roles": [
          "repository_context",
          "instruction_context"
        ],
        "basis": "User instructions and verified local root/HEAD/change list."
      },
      {
        "kind": "skill",
        "id": "agensi:promptbp-repository-pr-review@0.3.0",
        "roles": [
          "repository_review"
        ],
        "basis": "Actual Agensi creator-test delivery contains exact approved SKILL.md text."
      }
    ],
    "missing": [],
    "applied": false
  },
  "directive": {
    "role": "Read-only reviewer of the existing scope-fix patch",
    "objective": "Assess exact scope narrowing, preservation of operator context, out-of-scope HALT behavior and receipt compatibility.",
    "inputs": [
      "promptbp/environment.py",
      "promptbp/spine.py",
      "tests/test_review_skill.py",
      "Current HEAD plus unchanged working patch"
    ],
    "output_format": "Scoped findings, coverage, actual focused-test results, limitations and hash-bound executor receipt.",
    "performance_rules": [
      "Preserve all target content and existing edits.",
      "Run only the inspected focused tests; their file writes are confined to temporary directories outside the repository.",
      "Do not invoke the legacy preparation helper as the agent edition's implementation.",
      "No external writes, purchases, installs, publication, settings changes or source fixes."
    ],
    "style": "plain and concise",
    "recursive_check": [
      "All three changed files reviewed",
      "Target HEAD and working patch preserved",
      "Test results recorded accurately",
      "Review conclusion separated from release or merge approval"
    ]
  },
  "constraints": [
    "Scope is the current local patch only, not a complete new review of every historical package.",
    "Keep evidence outside the target.",
    "Stop if the source changes or a required check fails."
  ],
  "acceptance_criteria": [
    "Target HEAD and patch identified",
    "All three changed files inspected",
    "Focused regression suite passed",
    "Existing source/status preserved",
    "Result and receipt bound to actual saved output bytes"
  ],
  "evidence_required": [
    "Before/after HEAD, patch and source hashes",
    "Actual focused-test stdout/stderr",
    "Revision and file-line references",
    "Live connector delivery match"
  ],
  "gate": {
    "status": "ALLOW",
    "reasons": [
      "Scoped read-only validation and scratch-only test effects are authorized and available."
    ],
    "handoff_ready": true
  },
  "task_execution_status": "not_started",
  "next_action": "Execute the read-only review with the connector-delivered review skill.",
  "head_sha": "729d82fbdbf0692f368d3cc1b68b9c68bdd6e8d9",
  "working_patch_sha256": "f7d56807f84fe47b72aea190a94e92a6a61ff48001f46aee8f4bebe63d0a1977"
}

The separately authorized subsequent review passed 11 focused regression tests and preserved 141 source-file hashes, HEAD, status and the existing working patch. That execution has its own result and receipt; it does not change the plan's original not_started status.