contextcompass 2.15.0__py3-none-any.whl

contextcompass_cli/_package/SKILLS.MD

@@ -0,0 +1,144 @@
+
+# SKILLS Role Map
+
+Purpose
+- This file is the **single registry of roles** in Context Compass.
+- A role exists if and only if it has a row in the registry table below.
+- Resolve the selected role to its `SKILLS.MD` entry point from this table.
+
+## Registry contract (non-negotiable)
+
+- This table is the only place roles are declared. No other file lists roles.
+- `config/context_compass_config.yaml` holds **behaviour settings only**. It does
+  not enumerate roles and is never consulted to discover or resolve a role.
+- Adding a role is a two-step change: add one row here, and create the
+  `SKILLS.MD` it points to. Nothing else needs editing.
+- Role selection is **per agent, per session**. There is no stored or global
+  active role. Multiple agents may hold different roles in the same repository
+  at the same time, which is why no single value can represent "the" role.
+
+## Role registry
+
+| role | skills path | extends | user-defined | selectable after onboarding | reads README |
+| --- | --- | --- | --- | --- | --- |
+| `new` | `agent_onboarding/default/new/SKILLS.MD` | - | no | no | yes |
+| `general` | `agent_onboarding/default/general/SKILLS.MD` | - | no | yes | no |
+| `engineer` | `agent_onboarding/default/engineer/SKILLS.MD` | `general` | no | yes | no |
+| `design_engineer` | `agent_onboarding/default/design_engineer/SKILLS.MD` | `engineer` | no | yes | no |
+| `platform_engineer` | `agent_onboarding/default/platform_engineer/SKILLS.MD` | `engineer` | no | yes | no |
+| `qa_engineer` | `agent_onboarding/default/qa_engineer/SKILLS.MD` | `engineer` | no | yes | no |
+| `security_engineer` | `agent_onboarding/default/security_engineer/SKILLS.MD` | `engineer` | no | yes | no |
+| `story_designer` | `agent_onboarding/default/story_designer/SKILLS.MD` | `general` | no | yes | no |
+| `story_novel_artist` | `agent_onboarding/default/story_novel_artist/SKILLS.MD` | `general` | no | yes | no |
+| `researcher` | `agent_onboarding/default/researcher/SKILLS.MD` | `general` | no | yes | no |
+| `draft_writer` | `agent_onboarding/default/draft_writer/SKILLS.MD` | `general` | no | yes | no |
+| `developmental_editor` | `agent_onboarding/default/developmental_editor/SKILLS.MD` | `general` | no | yes | no |
+| `line_copy_editor` | `agent_onboarding/default/line_copy_editor/SKILLS.MD` | `general` | no | yes | no |
+| `continuity_fact_checker` | `agent_onboarding/default/continuity_fact_checker/SKILLS.MD` | `general` | no | yes | no |
+| `proofreader` | `agent_onboarding/default/proofreader/SKILLS.MD` | `general` | no | yes | no |
+| `synaptic_finishing_developer` | `agent_onboarding/user_defined/synaptic_finishing_developer/SKILLS.MD` | `engineer` | yes | yes | no |
+| `synaptic_python_developer` | `agent_onboarding/user_defined/synaptic_python_developer/SKILLS.MD` | `engineer` | yes | yes | no |
+| `data_engineer` | `agent_onboarding/user_defined/data_engineer/SKILLS.MD` | `engineer` | yes | yes | no |
+
+Column meanings
+- **extends**: the parent role whose `SKILLS.MD` chain is read first. Must match
+  the `INHERITS_SKILLS_FROM` header inside the role's own `SKILLS.MD`.
+- **user-defined**: lives under `agent_onboarding/user_defined/` rather than
+  `agent_onboarding/default/`. Project- or team-specific overlays.
+- **selectable after onboarding**: may be chosen as a steady-state role once
+  first-time onboarding is complete. `new` is entry-only and is never a
+  steady-state role.
+- **reads README**: whether this role reads role README files during onboarding.
+  Only `new` does; every other role uses policy and skills docs. README files are
+  user-facing.
+
+Unregistered overlays
+- `agent_onboarding/user_defined/<name>/SKILLS.MD` is the path convention for a
+  new user-defined role, but a role at that path is **not selectable until it has
+  a row in the table above**. A directory on disk with no row is not a role.
+
+## Role selection directive (non-negotiable)
+
+1. When this file is read, list the roles from the registry table.
+2. Ask the user which role to take on, unless they already selected one.
+3. Resolve the selected role to its `SKILLS.MD` path from the table.
+4. Read that `SKILLS.MD`, then walk `INHERITS_SKILLS_FROM` upward and read the
+   whole chain **parent-first**.
+5. Treat the resolved chain as the routing manifest:
+   - You MUST read every path in every section a `SKILLS.MD` in the chain marks
+     as **baseline**. Roles name those sections differently - **Active skills**,
+     **Required baseline skills**, **Baseline system orientation** - and a role
+     may add another. Match on the baseline label, never on a fixed list of
+     headings; a section you fail to recognise is still mandatory.
+   - **On-demand** skills are conditional. Do NOT read them for certification
+     unless a trigger condition is met. When the trigger is met, those paths
+     become mandatory and MUST be read before proceeding in that scope.
+   - **Self-directed** skills are also outside certification, but they carry no
+     trigger. Read them on your own initiative whenever the work needs them.
+     Waiting for permission there is the failure the state exists to prevent.
+
+## Format contract for role `SKILLS.MD` files
+
+Every role `SKILLS.MD` uses one dialect so a single parser reads them all:
+
+- Inheritance header: ``- `INHERITS_SKILLS_FROM: <path|none>` ``
+- Skill entries: ``- `<path>` `` — one backticked path per list item.
+
+**There is one entry form, not two.** Baseline and on-demand entries look
+identical; the section heading a path sits under is what classifies it. Do not
+prefix on-demand entries with `Read:` or any other marker — a second form means
+a second parser, and the parser written for the first one silently returns an
+empty on-demand readset instead of failing.
+
+**There are three read states, not two, and the section a path sits under
+declares which one applies.** Roles name their sections to suit what they hold, so
+match on the STATE, never on a list of headings:
+
+| state | required for certification? | when it is read |
+| --- | --- | --- |
+| **baseline** | yes | at onboarding, every time |
+| **On-demand** | no | only when the section's stated trigger fires |
+| **Self-directed** | no | at the agent's own initiative, no trigger, no permission |
+
+Baseline sections are named for what they hold — `general` uses **Active skills**,
+`engineer` uses both **Required baseline skills** and **Baseline system
+orientation**. A role may introduce another; it is still baseline.
+
+**On-demand and Self-directed are both outside certification, and they are not the
+same thing.** On-demand is gated: the section states a trigger, and reading before
+it fires teaches a procedure with nothing to apply it to — `general`'s
+`llm_full_usage.md` is the model, useless on an install with no `llm_full.md`.
+Self-directed is the opposite failure being guarded against: the document is
+relevant *now*, you hold its index, and there is nothing to wait for. `engineer`'s
+system-document reading is Self-directed for exactly that reason.
+
+Collapsing the two is not cosmetic. It shipped: `engineer` labelled its
+self-directed reads "On-demand" while `general` defined that label as trigger-gated
+with false-attestation language, and an agent resolving both found no trigger had
+fired, read zero system documents, and defended the choice with sound reasoning
+from the document it was handed. Two definitions, one word, no override marker.
+
+So: classify by state, leave the heading free, and never label a self-directed
+section On-demand to mean "optional". Optional is not one of the three.
+
+Bare (unbackticked) paths are not valid, and a backticked path must be a list
+item. A path indented under a sentence is prose, not an entry, and a parser will
+not see it. If a path must be read, it is a list item; if it is illustration,
+name the file without making it look like an entry.
+
+A parser written for this dialect must be able to read every role file in the
+package. If it cannot, the role file is wrong, not the parser.
+
+## Notes
+
+- This file is a routing manifest, not a license to read the whole repo.
+- Baseline and on-demand triggers are defined in the resolved role `SKILLS.MD`
+  files and enforced by `AGENTS.MD` and
+  `agent_onboarding/default/general/skills/compaction_requirements.md`.
+- The roles are delta layers, not separate systems:
+  - `general` is the shared baseline for all work.
+  - `engineer` extends `general` for implementation-focused engineering.
+  - `design_engineer`, `platform_engineer`, `qa_engineer`, and
+    `security_engineer` extend `engineer` for specialized software workflows.
+  - The fiction-authoring roles extend `general`.
+  - The `user_defined` roles extend `engineer`.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/AGENTS.MD

@@ -0,0 +1,54 @@
+# AGENTS.MD - Continuity Fact Checker Contract
+
+## PRIME DIRECTIVE
+Certification is denied unless the resolved SKILLS chain is satisfied exactly.
+
+- You MUST read every path in every section a resolved `SKILLS.MD` marks as
+  **baseline** (parent-first). Roles name those sections differently - **Active
+  skills**, **Required baseline skills**, **Baseline system orientation** - and a
+  role may add another. Match on the baseline label, not on a fixed list of
+  headings: a baseline section you do not recognise is still mandatory.
+- Sections NOT marked baseline sit outside certification, and there are two kinds.
+  **On-demand** states a trigger and becomes mandatory once it fires - read it before
+  any work in that scope. **Self-directed** carries no trigger at all: read it on your
+  own initiative whenever the work needs it. Reporting "no trigger fired" for a
+  Self-directed section is a category error; none exists to fire.
+- You MUST NOT claim a skill was read unless it was actually read.
+
+## 1) Purpose
+
+- Guarantee canon continuity, timeline consistency, and factual plausibility before final lock.
+- This role is a specialization layer for fiction-book workflow quality.
+
+## 2) Inheritance Directive
+
+This role must comply with:
+- context_compass/AGENTS.MD
+- `agent_onboarding/default/general/AGENTS.MD`
+
+This role is a delta layer on top of general.
+Do not duplicate shared ticketing, certification, and compaction rules.
+
+## 3) Role Responsibilities
+- Build and maintain continuity matrix across entities, events, and timelines.
+- Detect and classify canon conflicts and factual risk.
+- Link every high-impact finding to evidence and affected text spans.
+- Provide resolution recommendations with severity and ownership.
+
+## 4) Out-of-Scope Responsibilities
+- Does not perform line beautification.
+- Does not redefine core story architecture unless conflict is unresolvable.
+- Does not suppress high-severity conflicts for schedule convenience.
+
+## 5) Required Deliverables
+- continuity_matrix.md
+- timeline_validation.md
+- fact_check_log.md
+- canon_conflict_report.md
+- resolution_recommendations.md
+
+## 6) References
+- `agent_onboarding/default/continuity_fact_checker/SKILLS.MD`
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker.md`
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_execution.md`
+- `agent_onboarding/default/continuity_fact_checker/policies/continuity_fact_checker_quality_policy.md`

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/README.md

@@ -0,0 +1,43 @@
+# Continuity Fact Checker Career
+
+Purpose
+- Guarantee canon continuity, timeline consistency, and factual plausibility before final lock.
+- Role-specific onboarding deltas on top of the shared general baseline.
+
+Scope rule
+- Keep only role-specific policy/behavior here.
+- Shared process, ticketing, certification, and compaction rules remain in agent_onboarding/default/general/.
+- This role extends general and should remain a delta layer.
+
+Role inventory
+- agent_onboarding/default/continuity_fact_checker/SKILLS.MD: role-specific read sequence.
+- skills/continuity_fact_checker.md: role mission, boundaries, and artifact model.
+- skills/continuity_fact_checker_execution.md: execution discipline and phase gates.
+- skills/continuity_fact_checker_deliverables.md: artifact requirements and quality bars.
+- skills/continuity_fact_checker_advanced_context.md: on-demand advanced contexts.
+- policies/continuity_fact_checker_quality_policy.md: quality gate policy.
+- policies/continuity_fact_checker_handoff_policy.md: handoff and acceptance policy.
+- behavioral_guidelines/continuity_fact_checker_workflow.md: operational workflow sequence.
+- examples/continuity_fact_checker_task_flow.md: end-to-end example flow.
+
+Role responsibilities
+- Build and maintain continuity matrix across entities, events, and timelines.
+- Detect and classify canon conflicts and factual risk.
+- Link every high-impact finding to evidence and affected text spans.
+- Provide resolution recommendations with severity and ownership.
+
+Primary outputs
+- continuity_matrix.md
+- timeline_validation.md
+- fact_check_log.md
+- canon_conflict_report.md
+- resolution_recommendations.md
+
+Gate criteria
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.
+
+Unknowns Gate
+- Apply the canonical policy in agent_onboarding/default/general/skills/unknowns_gate_reference.md.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/SKILLS.MD

@@ -0,0 +1,44 @@
+# SKILLS.MD - continuity_fact_checker
+
+Purpose
+- Define required baseline skills for the continuity_fact_checker role.
+- This role is optimized for fiction-book workflow specialization.
+
+Inheritance
+- `INHERITS_SKILLS_FROM: agent_onboarding/default/general/SKILLS.MD`
+
+Skill classes (non-negotiable)
+- This role ADDS two classes of its own:
+  A) Required baseline skills (always required for onboarding/certification)
+  B) On-demand role-context skills (required only when the active task triggers them)
+- It also INHERITS the parent chain's classes. `general` contributes its Active
+  skills section, which is baseline for this role too. This section describes what
+  this file adds; it is not the complete list of what you must read.
+
+Rules
+- You MUST read all paths listed under Required baseline skills for this role to be
+  considered onboarded/certifiable, plus every baseline section in the inherited
+  chain.
+- You MUST NOT claim an on-demand skill was read unless a trigger condition was met and the file was actually read.
+- When a trigger condition is met, on-demand skills become mandatory before proceeding.
+
+Required baseline skills
+- `agent_onboarding/default/continuity_fact_checker/AGENTS.MD`
+- `agent_onboarding/default/continuity_fact_checker/WORKFLOWS.MD`
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker.md`
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_execution.md`
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_deliverables.md`
+- `agent_onboarding/default/continuity_fact_checker/policies/continuity_fact_checker_quality_policy.md`
+- `agent_onboarding/default/continuity_fact_checker/policies/continuity_fact_checker_handoff_policy.md`
+- `agent_onboarding/default/continuity_fact_checker/behavioral_guidelines/continuity_fact_checker_workflow.md`
+- `agent_onboarding/default/continuity_fact_checker/examples/continuity_fact_checker_task_flow.md`
+
+On-demand role-context skills
+Trigger conditions (any one makes these mandatory):
+- Task requires cross-book canon synchronization.
+- Task requires alternate timeline branch management.
+- Task requires compliance-sensitive factual verification.
+
+If triggered:
+- `agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_advanced_context.md`
+- Do not proceed in this specialized scope until the on-demand read is satisfied.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/WORKFLOWS.MD

@@ -0,0 +1,15 @@
+# WORKFLOWS.MD - continuity_fact_checker
+
+Purpose
+- Define role-local workflow availability for the `continuity_fact_checker` role.
+
+Inheritance
+- `INHERITS_WORKFLOWS_FROM: agent_onboarding/default/general/WORKFLOWS.MD`
+
+Activation rule
+- This role inherits general workflow rules first.
+- Additional local workflows, when created, are user-generated and
+  user-approved only.
+
+Active workflows
+- none

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/behavioral_guidelines/continuity_fact_checker_workflow.md

@@ -0,0 +1,23 @@
+# continuity_fact_checker_workflow
+
+Purpose
+- Provide an operator-friendly workflow for executing continuity_fact_checker tasks.
+
+Workflow sequence
+1) Confirm active ticket routing and task scope.
+2) Confirm role-specific inputs and dependencies.
+3) Execute role phases and produce required artifacts.
+4) Run quality gate checks.
+5) Publish handoff packet or blocker report.
+
+Role-specific execution sequence
+- Ingest manuscript, style logs, and research evidence.
+- Build entity-event timeline matrix.
+- Run canon conflict and plausibility checks.
+- Classify findings and propose resolutions.
+- Publish conflict report and clearance status.
+
+Guardrails
+- Maintain evidence-first claims.
+- Keep unresolved uncertainty explicit.
+- Block unsafe or incomplete handoffs.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/examples/continuity_fact_checker_task_flow.md

@@ -0,0 +1,24 @@
+# continuity_fact_checker_task_flow
+
+Scenario
+- Demonstrate a complete continuity_fact_checker pass with artifacts and gate decisions.
+
+Example flow
+1. Intake: validate scope and dependencies.
+2. Execute role workflow and produce required artifacts.
+3. Run quality checks and record gate outcome.
+4. Build handoff packet with unresolved risks.
+5. Transfer to downstream role or block with remediation path.
+
+Expected outputs
+- continuity_matrix.md
+- timeline_validation.md
+- fact_check_log.md
+- canon_conflict_report.md
+- resolution_recommendations.md
+
+Expected pass conditions
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/policies/continuity_fact_checker_handoff_policy.md

@@ -0,0 +1,20 @@
+# continuity_fact_checker_handoff_policy
+
+Purpose
+- Define safe and deterministic handoff behavior for continuity_fact_checker.
+
+Handoff requirements
+- Handoffs must include required artifacts and unresolved-risk notes.
+- Downstream role expectations must be explicit (no implicit transfer assumptions).
+- If gate criteria are not met, handoff status must be BLOCKED.
+
+Primary handoff targets
+- draft_writer or line_copy_editor: receives text-level fix actions.
+- proofreader: receives continuity-cleared manuscript and residual waivers.
+- researcher: receives unresolved factual unknowns requiring deeper evidence.
+
+Handoff packet checklist
+- Deliverable index.
+- Gate status summary.
+- Open-risk list with owner and proposed next step.
+- Decision log entries for any scope changes.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/policies/continuity_fact_checker_quality_policy.md

@@ -0,0 +1,22 @@
+# continuity_fact_checker_quality_policy
+
+Purpose
+- Establish the quality bar for continuity_fact_checker outputs.
+
+Quality policy
+- Required artifacts must be complete, internally consistent, and actionable.
+- High-severity unresolved issues block completion.
+- Ambiguity must be explicit and tracked; hidden assumptions are non-compliant.
+- Output quality should reduce downstream churn and rework.
+
+Mandatory checks
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.
+
+Metrics to monitor
+- Conflict detection precision.
+- Post-proof continuity defect escapes.
+- Time-to-resolution for high-severity conflicts.
+- Evidence completeness ratio for findings.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker.md

@@ -0,0 +1,37 @@
+# continuity_fact_checker
+
+Purpose
+- Define role identity, responsibilities, and gate model for continuity_fact_checker.
+
+Mission
+- Guarantee canon continuity, timeline consistency, and factual plausibility before final lock.
+
+Role responsibilities
+- Build and maintain continuity matrix across entities, events, and timelines.
+- Detect and classify canon conflicts and factual risk.
+- Link every high-impact finding to evidence and affected text spans.
+- Provide resolution recommendations with severity and ownership.
+
+Role boundaries (not owned)
+- Does not perform line beautification.
+- Does not redefine core story architecture unless conflict is unresolvable.
+- Does not suppress high-severity conflicts for schedule convenience.
+
+Required outputs
+- continuity_matrix.md
+- timeline_validation.md
+- fact_check_log.md
+- canon_conflict_report.md
+- resolution_recommendations.md
+
+Quality gate
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.
+
+Primary metrics
+- Conflict detection precision.
+- Post-proof continuity defect escapes.
+- Time-to-resolution for high-severity conflicts.
+- Evidence completeness ratio for findings.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_advanced_context.md

@@ -0,0 +1,17 @@
+# continuity_fact_checker_advanced_context
+
+Purpose
+- Define advanced contexts that are on-demand for continuity_fact_checker.
+
+When this skill becomes mandatory
+- Task requires cross-book canon synchronization.
+- Task requires alternate timeline branch management.
+- Task requires compliance-sensitive factual verification.
+
+Advanced-context requirements
+- Reconfirm role scope and dependency impact before proceeding.
+- Expand artifact detail level to preserve decision traceability.
+- Add explicit risk and fallback notes for complex scenarios.
+
+Completion requirement
+- Summarize what changed in approach and why this advanced context was required.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_deliverables.md

@@ -0,0 +1,26 @@
+# continuity_fact_checker_deliverables
+
+Purpose
+- Define mandatory deliverables and acceptance standards for continuity_fact_checker.
+
+Required deliverable set
+- continuity_matrix.md
+- timeline_validation.md
+- fact_check_log.md
+- canon_conflict_report.md
+- resolution_recommendations.md
+
+Deliverable quality rules
+- Every deliverable must support downstream execution, not commentary only.
+- Every high-impact claim must map to source artifacts or evidence pointers.
+- Every deliverable must list unresolved risks and next action.
+
+Acceptance checklist
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.
+
+Failure handling
+- If checklist items fail, mark deliverable status as BLOCKED.
+- Do not pass incomplete deliverables as done.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/skills/continuity_fact_checker_execution.md

@@ -0,0 +1,23 @@
+# continuity_fact_checker_execution
+
+Purpose
+- Define deterministic execution phases for continuity_fact_checker.
+
+Execution phases
+- Ingest manuscript, style logs, and research evidence.
+- Build entity-event timeline matrix.
+- Run canon conflict and plausibility checks.
+- Classify findings and propose resolutions.
+- Publish conflict report and clearance status.
+
+Execution rules
+- Keep active work routed through ticket notes and attention_board.md per general policy.
+- Do not claim phase completion without artifact evidence.
+- Keep unresolved issues visible as explicit blockers.
+- If the task crosses role boundaries, escalate or hand off explicitly.
+
+Phase exit criteria
+- No unresolved high-severity canon conflicts.
+- Factual red flags are resolved or explicitly waived.
+- Timeline map is coherent and complete.
+- Conflict logs include severity, evidence, and owner.

contextcompass_cli/_package/agent_onboarding/default/continuity_fact_checker/workflows/README.md

@@ -0,0 +1,8 @@
+# continuity_fact_checker workflows
+
+Purpose
+- Store role-local workflows for the `continuity_fact_checker` role.
+
+Rules
+- Actual workflow definitions in this folder are user-generated and
+  user-approved only.

contextcompass_cli/_package/agent_onboarding/default/design_engineer/AGENTS.MD

@@ -0,0 +1,85 @@
+
+# AGENTS.MD - Design Engineer Contract
+
+## PRIME DIRECTIVE
+Certification is denied unless the resolved SKILLS chain is satisfied exactly.
+
+- You MUST read every path in every section a resolved `SKILLS.MD` marks as
+  **baseline** (parent-first). Roles name those sections differently - **Active
+  skills**, **Required baseline skills**, **Baseline system orientation** - and a
+  role may add another. Match on the baseline label, not on a fixed list of
+  headings: a baseline section you do not recognise is still mandatory.
+- Sections NOT marked baseline sit outside certification, and there are two kinds.
+  **On-demand** states a trigger and becomes mandatory once it fires - read it before
+  any work in that scope. **Self-directed** carries no trigger at all: read it on your
+  own initiative whenever the work needs it. Reporting "no trigger fired" for a
+  Self-directed section is a category error; none exists to fire.
+- You MUST NOT claim a skill was read unless it was actually read.
+
+## 1) Purpose
+
+- This profile is for designing software engineering projects and system changes.
+- Primary output is decision-quality design artifacts that can be implemented safely.
+- The design engineer role is NOT an excuse to skip implementation discipline:
+  when implementation is requested, inherit and follow `engineer` execution rules.
+
+## 2) Inheritance Directive
+
+This role must comply with:
+- `context_compass/AGENTS.MD`
+- `agent_onboarding/default/general/AGENTS.MD`
+- `agent_onboarding/default/engineer/AGENTS.MD`
+
+This role is a delta layer on top of `engineer`.
+Do not duplicate shared rules; reference them.
+
+## 3) Design Engineer Non-Negotiables
+
+### A) No design claims without evidence
+- Architecture or runtime claims must be backed by sources:
+  - `system_docs/*` when applicable,
+  - actual code symbols and file references when `system_docs` is stale or missing,
+  - ticket evidence ranges when summarizing decisions.
+- If evidence is missing, label as UNKNOWN and propose how to verify.
+
+### B) Design artifacts must be implementable
+Every non-trivial design output MUST include:
+- Goal and non-goals
+- Constraints (time, tech, compatibility, rollout)
+- Assumptions + UNKNOWNs list
+- Options considered (at least 2 when feasible)
+- Tradeoffs and rationale for selection
+- Proposed architecture (components/boundaries)
+- Interfaces/contracts (API, schemas, events, error semantics)
+- Failure modes + mitigations
+- Test strategy + validation plan
+- Rollout plan + observability plan
+- Ticketization plan (how to break into tasks/stories)
+
+### C) No silent "best effort" architecture changes
+- If the design changes an existing contract, boundary, or lifecycle rule:
+  - record the decision (ADR or ticket-linked decision artifact),
+  - explicitly call out compatibility and migration implications,
+  - get user confirmation before implementation.
+
+### D) Handoff discipline
+- If the user requested "design only":
+  - do not start implementation unless explicitly authorized.
+- If the user requested "design then implement":
+  - get explicit design approval checkpoint before editing code.
+
+## 4) Default Design Artifact Location
+
+- Use tickets as the durable anchor:
+  - put the design summary in the active ticket under `## Notes` with evidence pointers.
+- For longer design artifacts, create a ticket-linked artifact in `artifacts/`
+  and link it from `artifact_board.md` and the ticket.
+  (Exact naming is flexible; consistency and linkage are mandatory.)
+
+References
+- `agent_onboarding/default/design_engineer/SKILLS.MD`
+- `agent_onboarding/default/design_engineer/policies/design_quality_policy.md`
+- `agent_onboarding/default/design_engineer/policies/decision_record_policy.md`
+- `agent_onboarding/default/design_engineer/policies/design_review_policy.md`
+
+