Skip to content

Topics

9 topic pages. A rule points to a topic when the detail is only needed in some situations.

Mechanically checked rules — details

Detail page for the rules that name topics/checks.md. A rule in .act/rules/ says what is forbidden or required; this page adds, per check, what it blocks, how it is switched and the limits an owner must know. How a check reads a command (parser allow-lists, time budgets, failure handling) is documented at its code in .act/hooks/checks/ (header comment of the module); the block messages name the syntax and point here. The checks are switched in docs/ai/config.md § Checks (block, warn or off).

Source: .act/rules/topics/checks.md

Project docs order — details

summary: what R-docs-order asks for, what act-doctor reports, how to switch it, how to move a document

Source: .act/rules/topics/docs-order.md

Feedback to the template author

Detail page for a rule that references it as topics/feedback.md. Read this when docs/ai/config.md § Feedback has feedback set to anything other than off, or when a template bug needs reporting regardless of that switch (see “Immediate trigger” below; with local, see “feedback: local”). Mechanism: .act/scripts/feedback.py (privacy checks in feedback_privacy.py), skill act-feedback.

Source: .act/rules/topics/feedback.md

Clickable entry ids

Detail page for the optional editor support that turns entry ids in Markdown files into links. Not loaded by a session start and tied to no rule: read it when a person wants ids such as T12 clickable in their editor. Nothing here applies to a project whose people do not install it.

Source: .act/rules/topics/id-links.md

IDE MCP server

Detail page for a rule that references it as topics/ide.md. Loaded only when a session start detects a connected IDE MCP server (JetBrains idea and similar names in .mcp.json, the Claude settings or ~/.claude.json, or Claude Code’s own ide server — see checks/session_context.py’s _ide_mcp_connected). Not loaded otherwise; nothing here applies to a project without one. Claude Code’s ide server offers getDiagnostics (read, fine for anyone) and executeCode (orchestrator only, like every tool that runs code).

Source: .act/rules/topics/ide.md

Ideas

Detail page announced by the session-start topic line. “ideas” is active when the session owner’s own ideas file — docs/ai/concept/ideas-<identity>.md, rules for the human in that folder’s README.md — has entries that are new or changed since they were last processed; the session start names them in an [act] ideas: note. Mechanism: .act/scripts/ideas.py, called from .act/hooks/checks/session_context.py.

Source: .act/rules/topics/ideas.md

Access to live systems

Detail page for R-safe-approval (.act/rules/shared/10-safety.md). Read this whenever a task reaches beyond the repo into a reachable system: a server over SSH, a database, a service’s API, a container host, a router, a smart-home or monitoring instance.

Source: .act/rules/topics/live-systems.md

Logging

Detail page announced by the session-start topic line. Read this when docs/ai/config.md § Logging has logging set to on — the session-start status line names every topic whose switch is on, this one included, see .act/hooks/checks/session.py’s refresh_session(). Mechanism: .act/scripts/log.py, observer .act/hooks/checks/event_log.py.

Source: .act/rules/topics/logging.md

Safeguard blocks

Detail page for R-safe-block (.act/rules/shared/10-safety.md). Read this whenever a tool flags a request or an action as risky — a guardrail, a content filter, a permission escalation.

Source: .act/rules/topics/safeguards.md