ccam80

Interactive top-down planning at thesis, chapter, section, paragraph, and sentence scope. Use to draft author-readable plan.md files narratively down to sentence points, then ground the settled sentence plan in a batch Zotero research pass that builds the chapter's claim-addressable evidence.md ledger for author haggling.

Derives the session's authorship tally from the conversation and the plan diff at session end. Presents draft for author approval before appending to the project's authorship_log.md, the only place authorship is recorded.

Conversational technical LaTeX writer. Use after an approved author-readable plan.md and matching evidence.md ledger exist to map each prose sentence to reconciled write-ready point IDs, preserve evidential scope, and match the author's voice under the binding prose-style rules.

Interactive top-down planning at thesis, chapter, section, paragraph, and sentence scope. Use to draft author-readable plan.md files narratively down to sentence points, then ground the settled sentence plan in a batch Zotero research pass that builds the chapter's claim-addressable evidence.md ledger for author haggling.

Derives the session's authorship tally from the conversation and the plan diff at session end. Presents draft for author approval before appending to the project's authorship_log.md, the only place authorship is recorded.

Conversational technical LaTeX writer. Use after an approved author-readable plan.md and matching evidence.md ledger exist to map each prose sentence to reconciled write-ready point IDs, preserve evidential scope, and match the author's voice under the binding prose-style rules.

Generates figures from placeholders in .tex files. Creates Python matplotlib scripts for data figures, TikZ/Python schematics for diagrams. Maintains a shared plot_defaults.py for consistent styling. Flags complex figures for the user. Runs after the writer skill.

LaTeX formatting checker and corrector. Ensures figure placement, equation alignment, table formatting, cross-references, and units comply with thesis conventions.

Execute implementation by driving a per-batch workflow that fans out implementers and verifiers, while the skill stays the human gateway for clarifications and user-required actions.

Review completed implementation against specs and rules. Drives a workflow that fans out one reviewer per phase, applies an internal auto-fix ruleset, presents only true decisions to the user, then applies fixes via a workflow.

Generate a high-level implementation plan with phases, waves, and verification measures. Collaboratively define goals, scope, and task structure with the user.

Take a phase from the implementation plan and produce a detailed implementation spec through collaborative architecting. Every design decision goes through the user.

Review phase specs for quality, consistency, and implementability before implementation. Drives a workflow that fans out one reviewer per phase and runs deterministic cross-phase checks in JS, then applies approved fixes via a workflow.

Spawnable research agent. Accepts bounded research requests, searches the indexed Zotero corpus through the deep-zotero MCP server, and returns claim-centred evidence cards with immediate verbatim passages, qualifications, contradictions, and search receipts. Callers spawn this via Task -- do not invoke directly. Never searches for, fetches, or imports external sources.