From ttal
Finalizes design plans via self-review, open questions identification, async external research, state updates, concise summaries, and automated review handoff with output channel partitioning.
npx claudepluginhub tta-lab/ttal-cli --plugin ttalThis skill uses the workspace's default tool permissions.
After a plan is written (by sp-planning or sp-debugging), this skill handles the final handoff phase: self-review, open questions, external research, state updates, and review routing.
Performs adversarial pre-coding planning: host and peer agents independently investigate codebase, diff specs, and validate plan with execution drill. Use for 'plan this', 'design approach', or scoping before coding.
Performs parallel multi-angle reviews of implementation plans from CEO, engineering, design, and devex perspectives. Merges scorecards, presents consolidated fix checklist for user selection, and applies approved changes.
Share bugs, ideas, or general feedback.
After a plan is written (by sp-planning or sp-debugging), this skill handles the final handoff phase: self-review, open questions, external research, state updates, and review routing.
Follow each step in order. Steps tagged (→ persist) write to state without surfacing to Neil. Steps tagged (→ human) output to Neil's context window.
Re-read the plan as the worker who has to execute it. For every step ask:
Note gaps to yourself only. Fix them in-place (edit the task tree / flicknote). No human-channel output from this step.
List decisions that genuinely need Neil's input (trade-offs you can't resolve, scope ambiguities, close architecture calls).
Write the list to persist first (annotate the task or append the orientation flicknote).
Then surface only the questions to the human channel — no reasoning, no alternatives, no self-review findings.
If there are no open questions, skip the human surface entirely.
If the plan needs cross-project reference, docs lookup, or web research, dispatch async:
ei ask --async "…" --project <alias>
ei ask --async "…" --web
Results land in ~/.einai/outputs/. Fold findings back into the plan. Do not narrate the research to Neil.
Apply everything from steps 1–3 directly to the task tree / flicknote / annotations. Silent persist work:
task <uuid> plan
flicknote modify <id>
task <uuid> annotate '<note>'
Surface only a concise summary of the finalized plan to Neil:
Keep it under ~200 words. No code blocks, no file paths, no step-by-step detail.
Run ttal go <uuid> to route the task to the plan reviewer. No human-channel output — the command itself is the handoff.
ttal go <uuid> ran cleanly in step 6Reminder: If persist-bound content leaked to the human channel, you burned Neil's context window.