agent-skills

Diagnose an agent FAILURE and prescribe the cheapest/earliest guardrail to add NEXT — then scaffold it ready to paste. Craft-on-the-fly, friction-driven, minimal: grow a harness to fit the real failure mode instead of designing one upfront. Sibling of /pick-model (picks the model) and /pick-workflow (picks the topology); this picks the GUARDRAIL. Domain-general — any agentic task, not just code. Use when 'agent keeps failing at X', 'it hallucinated/broke the contract/wasted tokens', 'what guardrail do I add', 'harness for X', 'how do I stop it doing Y', or when starting a new agentic task and you want a minimal starter harness. Recommends + scaffolds; does NOT run the task.

Resume session from CONTEXT-llm.md. Use when resuming work, loading saved context, continuing a previous session. Triggers include "load context", "resume session", "continue where I left off".

Decide HOW work should EXECUTE — linear vs parallel fan-out, sub-agents vs teams vs dynamic Workflow, and the per-step seam — instead of defaulting to linear+single-model. Produces an execution design, does NOT run the task. Delegates the per-step model+effort call to a model-picker skill when the environment has one. Use when authoring/challenging a skill or agent, AND mid-task when deciding how to run real work — 'should I fan this out', 'should this fan out', 'parallel or linear', 'sub-agents or workflow', 'per-step seam', 'execution topology', 'execution architecture'. For a trivial few-item call, early-exit to linear cheaply rather than a full analysis.

Save session to CONTEXT-llm.md with conversation summary. Use when saving work, checkpointing progress, preserving session state. Triggers include "save context", "save session", "checkpoint", "save my progress".

Challenge, push back, play devil's advocate on AI output or intentions. Use when: challenge this, are you sure, push back, prove it, what if you're wrong, devil's advocate, stress test, poke holes, second opinion, sanity check, too confident, really?, question this decision, avant de commencer, je veux faire X, aide-moi à décider. Modes: route (default — any invocation whose first word isn't a recognized subcommand, including a free-text description), forward (before starting), anchor (committed too fast), verify (facts wrong?), framing (wrong problem?), deep (full devil's advocate in fresh context).

Capture a session's verbatim "meat" — the exact phrasing of constraints, rejected paths, corrections, live reasoning, learnings, pivots, open questions — into CHECKPOINT-*-llm.md, plus a synthesized summary of state. Use before /clear on a long reasoning session, when exact wording matters, or when saving a session you'll resume cold. Triggers: "checkpoint", "save the meat", "capture verbatim", "checkpoint before clear".

Generate a copy-paste instruction block to steer Claude Code's /compact so the summary keeps what matters. Use when about to compact context intentionally, before a long session gets summarized, or when the user says "instruct compact", "compaction instructions", "steer compact", "custom compact", "prepare compaction".

McKinsey-style storyline builder for presentation decks. Turns a goal into a sequence of action-title slides with a logical problem→solution flow, where each line becomes one slide title. Use when structuring pitch decks, strategy decks, problem-solving or roadmap presentations, or when the user says "build a storyline", "structure my deck", "slide-by-slide outline", "action titles".

Evaluate whether the CURRENT model+effort fits a task, and recommend a switch only when worth the cache cost. Use when user asks "which model", "pick model", "model for", "what effort", "is haiku enough", "Opus or Sonnet", "should I downgrade/upgrade", or when routing a non-trivial task to a tier before starting costly/complex work. Judges model AND effort level. Does NOT execute the prompt — only judges fit. Covers tech and non-tech tasks.

Extract, define, and enforce a project's writing voice. Reverse-engineers a voice profile from real sample copy, or interviews you to build one, then applies it to drafts with a voice-constant / tone-flex model and validate-and-explain edits. Profiles are per-project artifacts at <project>/ref/tone-guide.md, resolved via config/projects.yaml. Use when user says "tone of voice", "voice guide", "brand voice", "make this sound like <project>", "extract voice", "apply voice", "does this match our tone".

Introspective end-of-session retrospective — the agent reflects on its OWN live context (method, decisions, dead-ends, token spend, lessons) and writes an honest, root-caused report ending in a reusable playbook. Use at the end of a long or complex session (30min+, multi-phase, sub-agents spawned, or a hard problem solved) when the user says "post-mortem", "retrospective", "session retro", "debrief this session", "what did we learn", "write up this session". NOT a log parser — it digs the current context, not JSONL files.

Creates, modifies, and orchestrates Claude Code tools (skills, agents, scripts). Use when user requests creating, updating, improving, or editing any Claude Code extension — skills, sub-agents, slash commands, script wrappers. Triggers include "create/make/new skill/command/agent/script", "tool for X", "slash command", "sub-agent", file paths with /skills/, /agents/. For plugin version bumps and release metadata, route directly to edit-plugin.

Audit, reorganize, and ensure consistency of ONE project folder against the zone+tier convention. Read-only audit → reorg plan → gated execute. Triggers: "sweep project", "tidy project", "reorganize project", "project hygiene", "audit folder", /sweep-project.

Recursive meta-prompt ledger for human-guided complex tasks spanning multiple sessions. Maintains RELIANCE.md — a single-source, append-aware file carrying FINALITÉ, the RAIL session-chain, and the decision/collaboration trail across context resets. Use when: a complex task spans sessions and you want loss-free hand-off + drift detection + proportional AI challenge, WITHOUT autonomous loops or heavy regression tests. Triggers: meta-prompt, reliance, start a guided task, resume reliance, close session, next RAIL.

Discover and evaluate Claude Code skills and plugins from curated sources. Use when searching for existing skills/plugins, finding inspiration, or before creating a new skill. Triggers include "search skill", "find skill", "find plugin", "skill for X", "what skills exist for", "discover skills".

Quick capture to GTD inbox from CLI or natural language

Convert Markdown files with Mermaid diagrams to styled PDF. Use when user wants to export/convert markdown to PDF, generate PDF documentation, or create printable documents.

Transcribe an audio recording (m4a/mp3/wav/ogg/flac) to markdown locally with whisper.cpp. Auto-detects language (multilingual, tuned for EN/FR), CPU-only, private — no API, no GPU. Self-bootstrapping: builds the pipeline on first use if missing. Use when the user wants to transcribe audio, convert a recording to text, get a transcript, or turn a voice memo into markdown. Triggers include "transcribe", "transcription", "audio to text", "voice memo to markdown".

Convert PDF to clean markdown. Auto-detects backend: pdftotext (fast, text-heavy) or docling (complex tables, multi-column, OCR). Use when converting PDFs to markdown for analysis or summarization. Triggers: "convert pdf", "pdf to markdown", "pdf file", "pdf to text", ".pdf".

Inline inbox triage — two-pass // comment flow for async Obsidian review. Triggers: process inbox, triage inbox, route inbox, process triage.

Autonomous multi-philosopher dialogue using Agent Teams. Use when: /encounter, autonomous dialogue, philosopher encounter, run philosophers, let philosophers talk.

Francis Fukuyama philosophical dialogue. Use when: /fukuyama, talk to Fukuyama, end of history, political order, institutional decay, trust, identity politics, liberal democracy, biotechnology, deep philosophy beyond coaching.

Frank Herbert philosophical dialogue. Use when: /herbert, talk to Herbert, ecology as politics, systems thinking, messianic trap, Dune philosophy, fiction as laboratory, bureaucracy, survival, deep philosophy beyond coaching.

Donella Meadows philosophical dialogue. Use when: /meadows, talk to Meadows, systems thinking, leverage points, feedback loops, limits to growth, sustainability, complex systems, system dynamics, deep philosophy beyond coaching.

Nassim Taleb philosophical dialogue. Use when: /taleb, talk to Taleb, antifragility, black swans, skin in the game, fat tails, risk, probability, fragility, uncertainty, deep philosophy beyond coaching.

Create baseline context from .in/ folder with manifest-driven organization (run once per project). Use when bootstrapping project context, setting up .ctx/ snapshot. Triggers include "create context", "bootstrap context", "setup context", "init context".

List all CONTEXT files across code/, projects/, repos/, and vaults/ with status and metadata. Use when viewing context registry, checking saved sessions, or syncing INDEX.md. Triggers include "list contexts", "show contexts", "context registry", "sync index".

Benchmark Praxis methodology against external frameworks, tools, or approaches. Use when: benchmark, compare praxis, how does praxis compare, gap analysis, benchmark vs, assess against, track changelog, watch features, inventory features, full feature sweep. Modes: full (default), quick, gap, track, inventory.

Divergent-convergent brainstorming with research, ideation, and trade-off analysis. Use when exploring options, generating ideas, comparing approaches. Triggers include "brainstorm", "explore options", "ideate", "compare approaches".

Chaotic domain: act to impose minimal constraints when cause-effect is invisible. Use when: production down with gibberish logs, completely novel problem, crisis, deliberate disruption, nothing makes sense. NOT for problems with any visible pattern — use /probe instead.

Sense-making before action. Classify problem using Cynefin triangulation (3 tests + decomposition) to route to the right skill chain. Use when: frame, what approach, how should I start, which skill, where to begin, unsure what to do. NOT for known tasks — just do them.

Deep proactive analysis for complex technical problems requiring upfront thinking and design. Use when: investigate, deep dive, technical spike, design strategy, complex multi-constraint problem, figure out how to, how should I approach. NOT for errors (use troubleshoot) or option brainstorming (use brainstorm).

Safe-to-fail experiment for Complex domain problems where cause-effect is only visible in retrospect. Two-phase: foreground qualify → background probe → sense result. Use when: probe, safe-to-fail, test hypothesis, experiment with hypothesis, Complex domain with hypothesis. NOT for brainstorming (use brainstorm) or known cause-effect (use investigate).

Use when user reports an error, bug, or something not working. Search-first troubleshooting with diagnostic phase. Triggers: debug, error, broken, not working, failing, crash, exception.

Detect and anonymize PII (SSN, cards, emails, phones, names) and business data (companies, revenue, costs, pricing). Use when handling sensitive files or user requests anonymization. Check-only or full anonymization with 5 strategies (mask/hash/pseudo/token/mixed). GDPR/HIPAA aware.

Capture cross-project connections on the fly. Use when you notice a link between two projects — a shared pattern, a people-bridge, a narrative that feeds another project. Persists to thinking/bridges/ as structured YAML. Use when user says "bridge", "this connects to", "this feeds into", "same pattern as", or explicitly names a cross-project link.

Convert markdown documents into AntV infographic SVGs. Analyzes content structure, selects optimal template, generates AntV syntax, and renders to SVG. Use when user wants to create infographics, visual summaries, or storytelling visuals from markdown.

On-demand code annotator for LLM context preservation. Adds section-level comments capturing intent, rationale, and gotchas. Use when "literatize", "add literate comments", "annotate this file", "add section comments".

Toshl Finance automation - sync data to local cache and generate monthly financial reports

Convert Word (.docx) files to clean markdown for LLM context (uses markitdown). Use when converting Word documents to markdown for analysis or summarization. Triggers include "convert docx", "word to markdown", "docx file".