
Record an explicit predicted observation before a side-effecting tool call. Use whenever you are about to run a command, edit a file, or invoke any tool whose outcome you have not already observed. Upholds the predict-before-act invariant.
Run a cheap, reversible real probe to ground a belief instead of hallucinating. Use when dreamer confidence is low or when you catch yourself guessing about repository state.
Surprise-triggered reflection. When a PostToolUse hook reports high prediction error (surprise >= 0.6), invoke /reflect to update beliefs via the belief-reviser and decide whether to continue, replan, reclarify, or ask the user.
Use when the user wants to add, remove, update, list, or sync plugins in this marketplace. Covers both remote (GitHub) and local plugin management, keeping marketplace.json and README.md in sync.
Cluster related instincts into higher-order skills, commands, or agents
Extract reusable patterns from the current session and save them as instincts
Analyze staged and unstaged changes and create logical, atomic git commits with auto-detected or user-specified commit message style
Gitea operations via gitea-mcp and tea CLI. Use when user mentions: gitea, tea, or when git remote shows a Gitea instance.
Register a new runbook with a name, trigger condition, and list of actions to automate repetitive task sequences
Manually execute a registered runbook by name, running its actions in sequence
Enable or disable a registered runbook by name - disabled runbooks skip automatic trigger execution
Bump the project version following semver, update all version files per https://semver.org/
Generate or update CHANGELOG.md using git-cliff from conventional commit history - see https://git-cliff.org/docs/category/usage
Validate version consistency across all project files, check semver format, git tag alignment, and changelog synchronization per https://semver.org/
Bootstrap versioning for a project - detect project type, initialize git-cliff config, and generate initial CHANGELOG.md using semantic versioning
List all registered runbooks with their triggers, actions, and enabled/disabled status
Export learned instincts for sharing or backup
Import instincts from an export file
Display all learned instincts with their confidence scores and domains
Create commits, push to remote, and open a PR/MR - auto-detects GitHub/Gitea/GitLab and skips PR if on the default branch
Create logical commits from current changes and push to remote - auto-detects GitHub/Gitea/GitLab and uses the appropriate CLI tool
Generate a detailed prompt for Claude Code by analyzing the codebase, then execute it immediately after user confirmation
Generate a detailed, context-rich prompt for Claude Code by analyzing the codebase and user intent - outputs the prompt for copying or saving
Delete a registered runbook by name from .claude/runbooks.json
Given a goal and a set of candidate next actions, rank them by Expected Free Energy (pragmatic + epistemic value) using past similar episodes, and pick the winner. Delegates scoring to the policy-selector subagent.
Snapshot the current belief state (and optionally git-stash the working tree) under a label, so a risky action can be rolled back if it produces high surprise or breaks acceptance tests.