cap_router_mgr

Router Rule Management

Use this skill to inspect and modify event router rules through the direct callable capabilities in cap_router_mgr.

When to use

• The task is to list, inspect, add, update, delete, or reload router rules.
• The task is to change router behavior by capability calls, not by editing rule files directly.
• The task requires deciding between add_router_rule and update_router_rule for one rule id.

Operating rules

• Use direct router manager capabilities, not console wrappers.
• Prefer changing exactly one rule per turn.
• Treat rule_json as a JSON string, not an embedded object.
• After add_router_rule or update_router_rule, always call get_router_rule.
• Use reload_router_rules only when the task explicitly requires disk reload semantics.

Capability contract

• list_router_rules Input: {} Output: JSON array of rule objects.
• get_router_rule Input: {"id":"rule_id"} Output: one rule object, or standard error JSON with ok=false.
• add_router_rule Input: {"rule_json":"{\"id\":\"rule_id\",\"match\":{\"event_type\":\"message\"},\"actions\":[{\"type\":\"drop\"}]}"}. Output: {"action":"add_router_rule","ok":true,"id":"rule_id"}. If the error says rule id already exists; use update_router_rule, switch to update_router_rule.
• update_router_rule Input: same shape as add_router_rule. Output: {"action":"update_router_rule","ok":true,"id":"rule_id"}. If the error says rule id not found; use add_router_rule, switch to add_router_rule.
• delete_router_rule Input: {"id":"rule_id"} Output: {"action":"delete_router_rule","ok":true,"id":"rule_id"}
• reload_router_rules Input: {} Output: {"action":"reload_router_rules","ok":true}

Standard workflow

1. Call list_router_rules.
2. Check whether the target id already exists.
3. Build one complete replacement rule object.
4. Call add_router_rule for a new id or update_router_rule for an existing id.
5. Call get_router_rule and verify the stored shape.

Rule contract

• Required fields in decoded rule_json:
  • id
  • match
  • actions
  • match.event_type
• Common optional rule fields:
  • description
  • enabled
  • consume_on_match
  • ack
  • vars
• Supported match fields:
  • event_type
  • event_key
  • source_cap
  • source_channel
  • chat_id
  • content_type
  • text
  • text_match_rule

Match semantics

• match.event_type is required; all other match fields are optional exact-match filters.
• match.text uses text_match_rule="exact" by default. Set text_match_rule="prefix" to match a leading command token, for example /run args.
• Prefix text matching trims leading whitespace from the event text, requires a token boundary after match.text, rejects embedded prose such as please /run args, and exposes the post-prefix argument string as {{match.remainder}}.
• event_type, event_key, source_cap, and source_channel values are not fixed enums in this skill; they must match values emitted elsewhere in the system.
• Use match.source_channel as the canonical rule field. The runtime also accepts channel as an alias.
• To run a script after device boot, match the startup trigger emitted by app_claw: source_cap=app_claw, event_type=startup, event_key=boot_completed, and content_type=trigger.

Actions

• Supported type:
  • call_cap
  • run_agent
  • run_script
  • send_message
  • emit_event
  • drop
• Optional action-level fields:
  • caller: system | agent | console
  • capture_output
  • fail_open
• Requirements and defaults:
  • call_cap: requires cap and input.
  • run_agent: input optional; defaults to {}. text falls back to event text. target_channel and target_chat_id fall back to the current event route.
  • run_script: requires input; common fields are path, args, async.
  • send_message: requires input; common fields are channel, chat_id, message. Missing channel falls back to target/source channel. Missing chat_id falls back to target endpoint/event chat id. Missing message falls back to previous action output.
  • emit_event: requires input; common fields are event_type, source_cap, source_channel, chat_id, message_id, content_type, text, payload_json, session_policy. Defaults: source_cap=claw_event_router, event_type=trigger, content_type=trigger, payload_json={}, invalid or missing session_policy=trigger.
  • drop: no special input.

Template context

• Action inputs are rendered against the current event context before execution.
• Common fields:
  • {{event.event_type}}
  • {{event.event_key}}
  • {{event.source_cap}}
  • {{event.channel}}
  • {{event.source_channel}}
  • {{event.target_channel}}
  • {{event.chat_id}}
  • {{event.message_id}}
  • {{event.content_type}}
  • {{event.session_policy}}
  • {{event.text}}
• Matched text fields are available when match.text matched:
  • {{match.text}}
  • {{match.rule}}
  • {{match.remainder}}
• Parsed payload fields are available under event.payload, for example {{event.payload.kind}}.

Decision policy

• If creating a new rule and the id already exists, do not call add_router_rule.
• If modifying an existing rule, inspect it first and preserve fields not meant to change.
• If it is unclear whether to create or replace, inspect first with list_router_rules or get_router_rule.
• Do not use reload_router_rules as a substitute for add or update.

Common failure causes

• Putting event_type at the rule top level instead of match.event_type.
• Passing rule_json as an object instead of a string.
• Omitting id or actions, or using an empty actions array.
• Using an unsupported action type.
• Mixing match.source_channel with action input channel.
• Using a relative path for run_script; use an absolute .lua path such as /fatfs/skills/.../scripts/action.lua.

Canonical examples

Run Lua script after boot

{
  "rule_json": "{\"id\":\"startup_run_lua_script\",\"description\":\"Run a Lua script once the device finishes booting.\",\"enabled\":true,\"consume_on_match\":false,\"ack\":\"startup Lua script executed\",\"match\":{\"source_cap\":\"app_claw\",\"event_type\":\"startup\",\"event_key\":\"boot_completed\",\"content_type\":\"trigger\"},\"actions\":[{\"type\":\"run_script\",\"input\":{\"path\":\"/absolute/path/to/your_script.lua\",\"args\":{}}}]}"
}

New session on /new

{
  "rule_json": "{\"id\":\"im_new_session\",\"enabled\":true,\"consume_on_match\":true,\"match\":{\"event_type\":\"message\",\"event_key\":\"text\",\"content_type\":\"text\",\"text\":\"/new\"},\"actions\":[{\"type\":\"call_cap\",\"cap\":\"new_chat_session\",\"input\":{}},{\"type\":\"send_message\",\"input\":{\"channel\":\"{{event.source_channel}}\",\"chat_id\":\"{{event.chat_id}}\",\"message\":\"Started a new session.\"}}]}"
}

Prefix command with arguments

{
  "rule_json": "{\"id\":\"im_run_script_command\",\"enabled\":true,\"consume_on_match\":true,\"match\":{\"event_type\":\"message\",\"event_key\":\"text\",\"content_type\":\"text\",\"text\":\"/run\",\"text_match_rule\":\"prefix\"},\"actions\":[{\"type\":\"run_script\",\"input\":{\"path\":\"/absolute/path/to/your_script.lua\",\"args\":{\"command\":\"{{match.remainder}}\"}}}]}"
}

Route IM text to agent

{
  "rule_json": "{\"id\":\"im_any_message_agent\",\"enabled\":true,\"consume_on_match\":true,\"match\":{\"event_type\":\"message\",\"event_key\":\"text\",\"content_type\":\"text\"},\"actions\":[{\"type\":\"run_agent\",\"input\":{\"target_channel\":\"{{event.source_channel}}\",\"session_policy\":\"chat\"}}]}"
}