dzackgarza

Turn a rough draft, vague ambition, or messy task brief into a powerful Codex /goal objective and, when needed, companion workflow docs for long-running autonomous work. Use when the user asks to write, improve, format, sharpen, stress-test, or activate a Codex goal, thread goal, durable goal, or /goal prompt.

HuggingFace hf CLI: search/download/upload models, datasets.

Decomposition playbook + specialist-roster conventions + anti-temptation rules for an orchestrator profile routing work through Kanban. The "don't do the work yourself" rule and the basic lifecycle are auto-injected into every kanban worker's system prompt; this skill is the deeper playbook when you're specifically playing the orchestrator role.

Pitfalls, examples, and edge cases for Hermes Kanban workers. The lifecycle itself is auto-injected into every worker's system prompt as KANBAN_GUIDANCE (from agent/prompt_builder.py); this skill is what you load when you want deeper detail on specific scenarios.

Use when reading a PDF — converts to markdown via Mistral OCR with local caching.

Use when you need to query Zotero data, find references, export citations, search for papers, or fetch PDFs using the local Zotero Web API cache.

Manage Zotero reference libraries via the Web API. Search, list, add items by DOI/ISBN/PMID (with duplicate detection), delete/trash items, update metadata and tags, export in BibTeX/RIS/CSL-JSON, batch-add from files, check PDF attachments, cross-reference citations, find missing DOIs via CrossRef, and fetch open-access PDFs. Supports --json output for scripting. Use when the user asks about academic references, citation management, literature libraries, PDFs for papers, bibliography export, or Zotero specifically.

Use when a hard task, repeated failure, unclear causality, or long-running goal creates pressure to pivot, defer, narrow scope, report blocked, or treat a failed one-shot attempt as an external blocker before recursive decomposition has reduced the residue.

Use when handling category-spec workflow state: cards, plans, decisions, TODO triage, retired cards, visual windows, branch/PR routing, validation handoff, or status updates.

Use when moving research work from plan to execution, deciding whether a card is atomic, routing implementation/self-check/adversarial audit, promoting or rejecting results, or claiming GOAL.md discharge.

Use when designing or reviewing category refinement order, Sage category creation, Hom/End/Aut structures, autsets, WithGenerators/FinitelyPresented axioms, dual-object routing, or constructor interception policy.

Use when doing post-hoc meta-review of completed cards, scanning for shallow/confabulated work that passed gate review, cleaning up planning debt, or identifying systemic Jerry-behaviour patterns across cards. Load when "planning cleanup", "meta-review", "completed card audit", or "card quality scan" is mentioned.

Use when auditing computational or formal proofs, writing or reviewing mathematical argument notes, checking evidence sufficiency, reviewing Sage/GAP/Lean/Aristotle verification, detecting proof fraud, or deciding whether a result is accepted.

Use after the user asks to plan category-spec work, decompose an approved plan into cards, or organize plan phases into task/decision/research work.

Use when editing, reviewing, or authoring category specs, type surfaces, Sage wrappers, constructors, method ownership, smoke files, or category-spec implementation code.

Use when working inside a specific category_specs subtree and needing local ownership, constructor, hom/end/aut, or test-directory guidance after nested AGENTS files were migrated into skills.

Agent behavior auditing and structural debugging. Use when diagnosing why agents fail, produce trivial work, or don't complete assigned tasks.

Documentation coverage work for lattice theory methods. Use when doing documentation-only assignments - adding/updating method documentation, checklists, or reference docs.

Interface and type-surface design work. Use when assignments define or change API contracts, type signatures, or behavioral guarantees.

Test coverage work for lattice methods. Use when assignments involve code, tests, or runtime behavior - writing or updating tests.

Debug autonomous agent workers, cron jobs, timeouts, and connectivity issues. Use when agent runs are failing, not running, or producing unexpected behavior.

Git commit message format and guidelines. Use when making commits or creating agents that make commits.

Use this skill to identify and eliminate deeply problematic semantic, structural, and lexical patterns that undermine high-quality writing. It provides a principle-based diagnostic guide for detecting statistical regression (the "Blurry Sketch" effect), intellectual dishonesty (bias by omission), adjectival inflation (puffery), and formulaic rhetorical contrast.