#SpeakUp Content Pipeline SOP

Status: Scaffold / triage findings. Not the SOP itself yet — that needs a dedicated session. This doc captures what exists, what's missing, and the open questions that need to be answered before the SOP can be written cleanly.

Why this exists: the pipeline has been built bottom-up across the LDH client work. Methodology files, skills, and crumbs of process exist scattered across the workspace. The SOP would consolidate them into a single document that says "for any new SpeakUp client, here's the end-to-end process, who runs each stage, what skills assist, what outputs land where, what review gates exist."

---

#Pipeline stages (current understanding)

From _ops/methodology/README.md + workspace-derived additions:

---

#Skill inventory mapped to stages

What exists today that could plug into each stage:

#Custom (workspace-built) skills

#Global skills relevant to the pipeline (NOT invoked this session — see open question)

#Skill gaps (candidates to build)

---

#Open questions before the SOP can be written

These are the things we don't have clean answers for. The SOP cannot be written until they're resolved (or explicitly deferred).

1. Cowork vs Code per stage. Some stages are heavy production runs (script writing, brief generation), others are deep methodology work (creative direction, engagement-DR design). The current feedback_tool_choice.md heuristic should be applied per-stage in the SOP. Which stages live where?

2. Skill invocation discipline. This session wrote 14 scripts without invoking copywriting, copy-editing, or brand-review. The Why-line bloat + em-dash slip would have been caught earlier with a brand-review pass. The SOP should specify "for stage X, invoke skill Y as soft-pass before delivering". Open: which skills are mandatory vs. optional per stage?

3. Soft-pass V2 protocol. Jason flagged the idea: take a draft, run it through a relevant skill on a different version, compare to original, decide which to ship. This needs a defined protocol: when to do it, how to file the V2, how to decide between versions. Candidate first test: take 3 LDH Mois 2 video scripts, run them through copywriting + brand-review, file as scripts-v2.md, compare.

4. Plankton routing. Stages 13-16 (production, publishing, engagement, measurement) live in Plankton, not here. The SOP needs to say where the Code-side handoff ends and Plankton picks up. Open in project_ldh.md.

5. Onboarding crumbs to consolidate. _ops/client-system/ONBOARDING.md + _ops/client-system/SPEC.md + clients/_template/README.md overlap. The SOP onboarding section should pull from all three, deduplicate, and become the single source.

6. Methodology vs skill split. Some pipeline knowledge lives as a _ops/methodology/*.md file (read by humans), some as a skill (executable). When does something become a skill vs stay a doc? Current rule of thumb: if a skill needs Plankton's framework registry to be useful (stories methodology), keep as doc until registry exists. Worth formalizing.

7. Per-client SOP variation. Pipeline is universal but clients have variations (LDH is the canonical, Cody/Wolf-Pack/etc. are legacy testing). Does the SOP have a "client variations" section, or does each client get its own SOP variant under clients/{slug}/?

8. Review gate convention. Between stages, what's the gate? (Draft posted to Notion? Slack message to Charles/Antonio? Brief committed to git?) Gates aren't documented today; happen ad-hoc.

---

#Soft-pass V2 protocol (next-session candidate)

The cleanest first test of skill-augmented pipeline work:

1. Take 3 LDH Mois 2 video scripts from scripts.md (varied: V1 tactical, V7 emotional, V9 humor test).
2. For each, run through:
   • copywriting skill (rewrite pass focused on hook strength + body tightening)
   • brand-review skill (compliance check against clients/les-dames-des-hypotheques/brand/messaging-framework.md)
3. File as scripts-v2.md in the same project folder.
4. Side-by-side comparison: did the skill-augmented version improve on the manual draft? Where? What did it miss that the manual draft caught?
5. Document the delta in a new _ops/methodology/skill-augmented-script-pass.md (or fold into this SOP scaffold).
6. If the V2 is meaningfully better, formalize the soft-pass as a mandatory SOP step for scripts. If not, document why and skip.

This is the cheapest way to learn whether the global marketing/copywriting skills add real value to the LDH-style work, or if the custom workspace methodology already covers what they offer.

---

#What gets built next session

Pickup order for the SOP-development session:

1. Read this file + _ops/methodology/README.md + _ops/client-system/ONBOARDING.md + _ops/client-system/SPEC.md.
2. Resolve open question #1 (Cowork vs Code per stage) and #6 (methodology vs skill split). These two unblock most of the rest.
3. Run the soft-pass V2 protocol on 3 LDH scripts. Decision on inclusion in standard flow.
4. Write the SOP itself: stage-by-stage table with owner, input, skills, output, gate. Replace this scaffold doc.
5. Resolve onboarding crumb consolidation (#5). Probably means _ops/client-system/ONBOARDING.md becomes the source, others reference.
6. Update CLAUDE.md client-system auto-behaviors to point to the new SOP.

---

#Cross-references

• _ops/methodology/README.md — pipeline order from Angelo + framework index
• _ops/client-system/ONBOARDING.md — current client onboarding
• _ops/client-system/SPEC.md — client folder spec
• clients/_template/ — folder template
• feedback_tool_choice.md (auto-memory) — Cowork vs Code heuristic
• feedback_deliverable_hygiene.md (auto-memory, new this session) — keep client deliverables free of internal reasoning