dryas

Plugin dryas from marketplace dryas-workflow in promprit/dryas-workflow. Commands read from the repo's manifest — 3 days ago.

Skills (12)

• benchmark-checklist

  Vet a performance number you measured (limiter, tuning, limits, errors, repeatability, relevance, whether the work ran) before you report or act on it. Runs from a PLAN.md Done line, the CLAUDE.md rule, or /benchmark-checklist.

• brainstorm

  Use before any creative or multi-step work (features, components, new behavior, changes to behavior). Establishes intent and gets the user's approval of a design on the right path (spike, bounded, architectural) before anything is built or dispatched. The Dryas "brainstorm first, always" gate.

• debug

  Use on any bug, test failure or unexpected behavior, before proposing a fix, and before /orchestrate climbs a task to a stronger tier. Root cause first (feedback loop, then four phases); no fix by guessing.

• interrogate

  Adversarial review of a finished change by two reviewers on different models, then a lead verdict. Runs from the orchestrate skill's gated interrogate step or when the user types /interrogate. Not for routine review; /wreview covers that.

• lens

  Use when a design needs review (UI, UX, layout, visual style, components, on-screen copy), when brainstorm meets a new feature or product idea, or when the user says "3 lens" or /lens. Runs the design panel (craft, impeccable, ui-ux-pro-max) or the product panel (focus, craft, data), returns a fixed

• orchestrate

  Run a multi-step coding task as strong-tier orchestrator with fast-first executors (standard, strong, frontier on escalation), Jev routing/gates, Dryas memory and the Dryas process skills. Use when the user runs /orchestrate, asks to orchestrate or dispatch executors, or has approved a multi-step pl

• plan

  Use when an approved design or spec needs an implementation plan, before touching code. Writes a plan of self-contained `## Task N:` sections (Goal, Scope, Done, Failing test first) that /orchestrate runs task by task. Used by /wplan.

• review-receive

  Use when acting on review findings, before changing anything. Verify each finding against the code, ask when it is unclear, push back with evidence when it is wrong, fix in order of severity; no performative agreement.

• review-request

  Use when a task or branch is done and before any merge. Requests a three-stage review of the diff from three fresh reviewers in parallel, spec compliance, code quality and over-engineering, with findings ranked by severity.

• tdd

  Use when implementing any feature or bug fix, executors included, before writing implementation code. Write the failing test first, watch it fail for the right reason, write the minimal code to pass, then refactor.

• verify

  Use before claiming work is done, fixed or passing, and before /commit or any merge. Run the command that proves the claim, read its full output, then state the result with the evidence; evidence before assertions, always.

• worktree

  Use when starting multi-step work or before running a plan. Creates or confirms the task's own git worktree and branch (one agent per worktree), sets it up, and checks a clean test baseline before any change.

Slash commands (9)

• /dryas:commit

  Conventional commit, gated by a Jev "safe to commit?" check

• /dryas:dryas-cloud

  Check, set up, update or remove the Dryas copy that makes this repo work in claude.ai/code cloud sessions

• /dryas:handoff

  Score this session with Jev and save a handoff; then /clear starts fresh with only what matters (/handoff move: for a session on another account)

• /dryas:lens

  Run the lens review (design panel or product panel) and return the `## Lens` block

• /dryas:orchestrate

  Run a multi-step task with the strong-tier orchestrator, fast-first executors and Jev gates

• /dryas:retro

  Session retrospective — propose environment improvements by severity; proposes, never applies

• /dryas:tune

  Weekly Jev calibration - override rates, judged share, frontier dispatches, swarm vs plain, spend; propose threshold changes

• /dryas:wplan

  Write an implementation plan in the Dryas plan format

• /dryas:wreview

  Three-stage review (spec compliance, code quality, over-engineering) of the current branch or worktree diff

Subagents (1)

• executor

  Implements exactly one task section of .orchestrate/PLAN.md under a locked file scope, test-first. Dispatched by the orchestrate skill with only its own section; not for open-ended work.

MCP servers (1)

• jev

  runs: ${DRYAS_PY:-python3} -X utf8 ${CLAUDE_PLUGIN_ROOT}/jev/jev_mcp.py

Hooks (14)

• PreToolUse

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" pretool

• UserPromptSubmit

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" prompt

• PostToolUse

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" post

• PostToolUseFailure

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" post

• SubagentStop

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" subagent-stop

• Stop

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" stop

• SessionEnd

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" session-end

• SessionStart

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/rules_inject.py"

• SessionStart

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" session-start

• SessionStart

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/compact_restore.py"

• SessionStart

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/handoff_restore.py"

• SessionStart

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/resources.py" --hook

• PreCompact

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/compact_keep.py"

• PreCompact

  "${DRYAS_PY:-python3}" -X utf8 "${CLAUDE_PLUGIN_ROOT}/jev/chain.py" precompact

Similar plugins