forgeplan_fpf_rules

Returns the active FPF rule set — the condition trees that drive Forgeplan’s action recommendations. Each rule has a name, a priority, an action bucket (EXPLORE / INVESTIGATE / EXPLOIT), a condition that evaluates against an artifact’s state, and a message explaining the resulting action. Rules are loaded from .forgeplan/config.yaml (fpf.rules) if present, otherwise from the built-in defaults (PRD-041).

Category: FPF Knowledge Base

When an agent calls this

“Which rules are currently in effect for this workspace?” — the agent wants an inventory before debugging a surprising action.
“Show me only the EXPLORE rules” — narrow the view when reasoning about low-trust artifacts.
“Is rule X from config or from defaults?” — use source to diagnose whether a custom override is actually loaded.
“Give me a one-line summary of every rule” — pass summary: true to skip the full condition trees.

Action buckets

Forgeplan’s engine recommends actions based on R_eff thresholds (configurable via fpf.thresholds in config). The three buckets are:

Bucket	Default R_eff range	Meaning	Typical rule effects
EXPLORE	`R_eff < explore_reff` (default 0.33)	Trust too low — treat as hypothesis	Add evidence, widen hypotheses, reduce scope
INVESTIGATE	`explore_reff ≤ R_eff < investigate_reff` (default 0.66)	Trust partial — test, measure	Run benchmarks, adversarial review, narrow hypotheses
EXPLOIT	`R_eff ≥ exploit_reff` (default 0.66)	Trust high — act	Ship, activate, supersede, enforce

Depth influences thresholds: higher depth (Deep / Critical) raises the bar for EXPLOIT, so a Critical artifact needs stronger evidence than a Tactical one to earn the same bucket.

Input parameters

Name	Type	Required	Description
`action`	`string`	no	Filter to one bucket: `"EXPLORE"` / `"INVESTIGATE"` / `"EXPLOIT"`. Omit for all.
`name`	`string`	no	Fetch a single rule by exact name. Errors if not found.
`summary`	`bool`	no	If `true`, returns only `{name, priority, action}` per rule — no condition trees, no messages. Default `false`.
`source`	`string`	no	Filter by origin: `"config"` (user overrides) or `"default"` (built-ins). Useful for debugging config loading.

Schema source: crates/forgeplan-mcp/src/server.rs::FpfRulesParams

Returns

Full mode (default):

{
  "source_in_use": "config",
  "count": 14,
  "rules": [
    {
      "name": "low_trust_explore",
      "priority": 10,
      "action": "EXPLORE",
      "condition": {
        "all": [
          { "field": "r_eff", "op": "lt", "value": 0.33 },
          { "field": "status", "op": "eq", "value": "draft" }
        ]
      },
      "message": "R_eff below explore threshold; add evidence before deciding.",
      "source": "config"
    }
  ]
}

Summary mode (summary: true):

{
  "count": 14,
  "rules": [
    { "name": "low_trust_explore",   "priority": 10, "action": "EXPLORE" },
    { "name": "stale_investigate",   "priority": 20, "action": "INVESTIGATE" },
    { "name": "high_trust_exploit",  "priority": 30, "action": "EXPLOIT" }
  ]
}

Example invocation

List all rules:

{}

Only EXPLORE rules, summary form:

{ "action": "EXPLORE", "summary": true }

Fetch a specific named rule:

{ "name": "high_trust_exploit" }

Debug: are my config overrides actually loaded?

{ "source": "config" }

Typical sequence

forgeplan_fpf_rules                  ← inventory
forgeplan_fpf_check { id: "PRD-19" } ← which rule wins for this artifact
forgeplan_fpf_section { id: "B.3" }  ← why the engine thinks that way

CLI equivalent

forgeplan fpf rules — same listing in the terminal.