wheakerd
@wheakerd
Public Skills
using-axiom
by wheakerd
Route startup, resume, compaction, requests and local web research. The startup Hook loads this gate; invoke it directly only on an explicit using-axiom request.
verify-completion
by wheakerd
Verify a completion, progress, or success claim against current direct evidence before it is made or accepted, especially after context compaction, resume, or a long task, and keep advice separate from compliance. Use when the user asks to verify or audit whether work is really complete, when a task resumes or continues after compaction with unfinished mutations, or when a completion claim rests on recalled or summarized facts. Classifies each fact as observed, recalled, or summarized, rebuilds a short evidence ledger for every completion criterion, re-derives the task state from files, commands, and records instead of summaries, and states advice and risks before complying when evidence conflicts with the request. Do not use for ordinary status questions, routine continuation of an unfinished task, reviewing an identified Axiom task's routing and authorization, or supervised multi-session progress.
agents-architect
by wheakerd
Design, initialize, audit, split, migrate, maintain, or validate a repository's AGENTS.md instruction system, routed .agents guidance, and repo-local .agents/skills or .claude/skills that support that system. Use for active-instruction discovery, durable AGENTS updates, route ownership, oversized guidance, instruction-memory growth, task-state placement, or an explicit effective-instructions, effective-instructions:preview, effective-instructions:refactor, effective-instructions:force, effective-instructions:reconcile, or effective-instructions:reconcile-preview mode. Do not use for packaged or installed plugin Skills, ordinary documentation, or general agent usage optimization unless the requested change is specifically to AGENTS architecture.
agent-plugin-architect
by wheakerd
Design, initialize, audit, migrate, maintain, or evaluate a packaged Codex or Claude Code plugin's architecture across shared Skills, route ownership, manifests, marketplace wrappers, hooks, and version-bound compatibility evidence. Use only for explicit packaged agent-plugin architecture work. Do not use for evaluating or reviewing an individual Skill or Skill proposal, repository-local AGENTS.md, .agents/skills, or .claude/skills systems, ordinary source-code or documentation work merely because it is in a plugin repository, host installation, publication, deployment, or Git submission. When architecture and installation or deployment are mutually exclusive alternatives, ask one concise clarification before routing and keep the workflow unselected until the user chooses; delegation or execution limits do not resolve that choice. Preserve each option's original end goal; state permission limits separately instead of replacing a goal with preparation, review, planning or preflight.
optimize-agent-usage
by wheakerd
Reduce or diagnose agent usage of tokens, credits, and context without weakening task quality or safety. Use only when the user explicitly asks to save credits, tokens, or usage limits in the current agent host; optimize Skills, instruction files such as AGENTS.md or CLAUDE.md, MCP, or context loading; explain unusually high usage; or design a lower-consumption workflow. Do not use for ordinary coding, software performance work, or generic cost questions.
clarify-intent
by wheakerd
Clarify a user request when its wording supports materially different intended outcomes that current context cannot resolve. Offer the most plausible options and accept a custom answer before dependent work. Use on an explicit invocation or when using-axiom selects this ambiguity boundary; do not trigger on tentative wording alone, routine implementation choices, or facts available through inspection.
reversible-system-change
by wheakerd
Plan, prepare recovery material, rehearse, or execute persistent system changes with verified rollback and postconditions. Use for an install, upgrade, deployment, migration, destructive retention action, active-version promotion, or explicit machine-credential lifecycle inventory, plan, activation, or retirement with rollback, data, service, or activation risk. Plans and non-mutating workflow rehearsals stay read-only; recovery preparation and isolated restore rehearsals require bounded write authority. Pair with confirm-external-action when the same request also causes a consequential external app or account effect. Do not use for ordinary source/configuration edits, Git operations, generic status queries, or conceptual explanations. When installation or deployment and architecture work are mutually exclusive alternatives, clarify before routing; do not select a different write scope on the user's behalf.
delegate-simple-task
by wheakerd
Assess clear, simple, bounded work using the user's ordered subagent candidates while preserving the main model. Invoke only through an enabled using-axiom gate with the host's verified top reasoning level and the user's confirmation or explicit authorization to invoke this skill. A disabled, absent, or unverified gate, a lower or unknown reasoning level, or missing invocation consent prevents activation, including direct requests and any permission-bypass mode. Keep invocation consent separate from assignment authority.
task-planning
by wheakerd
Create or revise an actionable task plan from the user's current requirements. Use when the user explicitly requests a task or implementation plan, or asks to remove, replace, merge, reorder, or narrow items in an existing plan. Keep retained decisions, dependencies, and acceptance criteria consistent with the current scope. Do not use for scheduled automations, progress reports, ordinary chat corrections, plan execution, or specialized planning already owned by another Axiom route. An explicit task-planning invocation may refine that plan's presentation while preserving its owner's constraints and authority.
traceable-git-submit
by wheakerd
Clarify unresolved material either/or choices before routing; delegation or checkpoint options do not select. Route an explicit invocation of this skill, checkpoint/baseline, consolidation/recovery, hardened/multi-target/history-replacing Git; combined commit, tag, and push of an already-prepared plugin release. Ordinary named-remote non-force Git without a tag or such needs and concepts stay host-native.
local-web-search
by wheakerd
Use before web search or follow-up page access from the user's machine or local workspace through a browser, CLI, script, or local client, including a fallback from cloud search. Also classify an uncertain execution location before that access. Confirmed cloud-hosted search and unrelated web app actions stay outside.
continuous-progress
by wheakerd
Supervised progress across host sessions by explicit invocation only. Use when the human invokes this skill to pursue a defined outcome through fresh execution sessions, coordinator handoffs, compaction, and independent recovery. Do not select it for ordinary requests to continue, one-off work, skill evaluation or authoring, or a request to use native Goals or /goal alone. Activation requires verified native scheduling, a supported independent supervisor, and existing authority for the requested coordination.
confirm-external-action
by wheakerd
Bind and verify a consequential external action. Use when the user explicitly asks to send, publish, invite, purchase, trade, delete, create or revoke a machine credential, or change external app or account state and the exact actor, target, payload, disclosure, cost, or retry boundary matters. Do not use for read-only lookup, draft-only work, or local Git. Pair with reversible-system-change when the same request also has persistent rollback, data, service, or activation risk. When review and execution are mutually exclusive alternatives, clarify the intended outcome before routing; a review request grants no execution authority.
review-axiom-task
by wheakerd
Review an Axiom task's routing, scope, authorization, actions, evidence, stops and outcome; explain, audit, criticize, appeal or narrow its decisions, including why it selected, allowed or refused an action. Ordinary summaries, code review, execution and retry stay outside. Clarify mutually exclusive review versus execution/retry before routing; review grants no action authority.