| 1 | package agent |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "encoding/json" |
| 6 | "fmt" |
| 7 | |
| 8 | "reasonix/internal/plancontract" |
| 9 | "reasonix/internal/tool" |
| 10 | ) |
| 11 | |
| 12 | // SubmitPlanTool is the planner's structured exit: it hands the host a plan as |
| 13 | // data instead of prose the host would have to parse back. Validation runs here, |
| 14 | // so a malformed plan returns an actionable tool error the planner can fix in |
| 15 | // the next round rather than a silent misparse downstream. |
| 16 | type SubmitPlanTool struct{} |
| 17 | |
| 18 | // finalizesTurn marks submit_plan as a host-consumed terminal tool. The marker |
| 19 | // is deliberately package-private: arbitrary plugin tools cannot opt into |
| 20 | // ending an Agent.Run without an owning host contract. |
| 21 | func (*SubmitPlanTool) finalizesTurn() {} |
| 22 | |
| 23 | func NewSubmitPlanTool() *SubmitPlanTool { return &SubmitPlanTool{} } |
| 24 | |
| 25 | func (*SubmitPlanTool) Name() string { return tool.HostSubmitPlan } |
| 26 | |
| 27 | func (*SubmitPlanTool) Description() string { |
| 28 | return "Submit your finished plan as structured data. This is how a plan reaches the host — the host renders it for the user and hands it to the executor, so do NOT also restate the plan in prose. Every step needs a `title`; a step with a `parent_id` is a sub-step of that phase (two levels, keep phases few). Record what you actually READ as `verified_files` and what you only INFERRED as `candidate_files` — never present a guess as a verified path. Attach `acceptance` criteria and command-level `verification` to the steps they belong to, mark must-keep-passing behavior with `regression`, and label anything unproven in `assumptions`. Set `requires_approval` when execution should stop for the user first; the host decides whether it actually gates." |
| 29 | } |
| 30 | |
| 31 | func (*SubmitPlanTool) Schema() json.RawMessage { |
| 32 | return json.RawMessage(`{ |
| 33 | "type":"object", |
| 34 | "properties":{ |
| 35 | "objective":{"type":"string","description":"What this plan achieves, in one sentence."}, |
| 36 | "assumptions":{ |
| 37 | "type":"array", |
| 38 | "description":"Premises the plan rests on that you did NOT verify. Label them here instead of stating them as facts.", |
| 39 | "items":{ |
| 40 | "type":"object", |
| 41 | "properties":{ |
| 42 | "text":{"type":"string","description":"The unverified premise."}, |
| 43 | "confirm":{"type":"string","description":"The cheapest check that would settle it."} |
| 44 | }, |
| 45 | "required":["text"] |
| 46 | } |
| 47 | }, |
| 48 | "non_goals":{"type":"array","items":{"type":"string"},"description":"Explicitly out of scope."}, |
| 49 | "steps":{ |
| 50 | "type":"array", |
| 51 | "minItems":1, |
| 52 | "description":"Ordered steps. Top-level steps are phases; give a step a parent_id to make it a sub-step of that phase.", |
| 53 | "items":{ |
| 54 | "type":"object", |
| 55 | "properties":{ |
| 56 | "id":{"type":"string","description":"Short id for this step, e.g. \"s1\". Only needed when another step references it."}, |
| 57 | "parent_id":{"type":"string","description":"The id of the phase this step belongs to. Omit for a phase."}, |
| 58 | "title":{"type":"string","description":"Imperative description of the step."}, |
| 59 | "depends_on":{"type":"array","items":{"type":"string"},"description":"Ids of sibling steps that must happen first."}, |
| 60 | "verified_files":{"type":"array","items":{"type":"string"},"description":"Paths you actually opened and read."}, |
| 61 | "candidate_files":{"type":"array","items":{"type":"string"},"description":"Paths you inferred but did NOT read. Never list an unread path as verified."}, |
| 62 | "acceptance":{ |
| 63 | "type":"array", |
| 64 | "description":"Checkable conditions this step must meet.", |
| 65 | "items":{ |
| 66 | "type":"object", |
| 67 | "properties":{ |
| 68 | "text":{"type":"string","description":"The condition, stated so it can be checked."}, |
| 69 | "regression":{"type":"boolean","description":"True when this is existing behavior that must keep working."}, |
| 70 | "optional":{"type":"boolean","description":"True for a nice-to-have that must never block completion."} |
| 71 | }, |
| 72 | "required":["text"] |
| 73 | } |
| 74 | }, |
| 75 | "verification":{ |
| 76 | "type":"array", |
| 77 | "description":"Commands that prove the step.", |
| 78 | "items":{ |
| 79 | "type":"object", |
| 80 | "properties":{ |
| 81 | "command":{"type":"string","description":"The command as the executor should run it."}, |
| 82 | "expect":{"type":"string","description":"What a pass looks like."} |
| 83 | } |
| 84 | } |
| 85 | }, |
| 86 | "risks":{"type":"array","items":{"type":"string"},"description":"What could go wrong in this step."} |
| 87 | }, |
| 88 | "required":["title"] |
| 89 | } |
| 90 | }, |
| 91 | "requires_approval":{"type":"boolean","description":"Request that execution stop for explicit user approval. The host owns the final decision."} |
| 92 | }, |
| 93 | "required":["objective","steps"] |
| 94 | }`) |
| 95 | } |
| 96 | |
| 97 | // ReadOnly is true: submitting a plan records a proposal and touches nothing. |
| 98 | func (*SubmitPlanTool) ReadOnly() bool { return true } |
| 99 | |
| 100 | // ProviderVisible gates on the host having armed a planning turn. The schema |
| 101 | // stays constant for cache stability and availability is decided when the call |
| 102 | // runs. |
| 103 | func (*SubmitPlanTool) ProviderVisible(ctx context.Context) bool { |
| 104 | _, ok := planSubmissionFromContext(ctx) |
| 105 | return ok |
| 106 | } |
| 107 | |
| 108 | func (*SubmitPlanTool) Execute(ctx context.Context, args json.RawMessage) (string, error) { |
| 109 | submission, ok := planSubmissionFromContext(ctx) |
| 110 | if !ok { |
| 111 | return "", fmt.Errorf("submit_plan is only available while planning; there is no plan to submit in this phase") |
| 112 | } |
| 113 | var plan plancontract.Plan |
| 114 | if err := json.Unmarshal(args, &plan); err != nil { |
| 115 | return "", fmt.Errorf("invalid args: %w", err) |
| 116 | } |
| 117 | plan = plan.Normalize() |
| 118 | if err := plan.Validate(); err != nil { |
| 119 | return "", fmt.Errorf("plan not accepted: %w", err) |
| 120 | } |
| 121 | plan = submission.record(plan) |
| 122 | phases, subSteps := 0, 0 |
| 123 | for _, step := range plan.Steps { |
| 124 | if step.ParentID == "" { |
| 125 | phases++ |
| 126 | continue |
| 127 | } |
| 128 | subSteps++ |
| 129 | } |
| 130 | approval := "" |
| 131 | if plan.RequiresApproval { |
| 132 | approval = " Approval was requested; the host decides whether execution gates." |
| 133 | } |
| 134 | // A revision is told what it changed. Left to describe its own edit a model |
| 135 | // reports intent, not effect, and a step it dropped by accident reads the |
| 136 | // same as one it kept. |
| 137 | revised := "" |
| 138 | if diff, ok := submission.Revised(); ok && diff.Moved() { |
| 139 | revised = "\n\n" + plancontract.RenderDiff(diff) |
| 140 | if diff.NeedsApproval() { |
| 141 | revised += "\n\nThis revision expands the approved scope; the host may gate it again." |
| 142 | } |
| 143 | } |
| 144 | return fmt.Sprintf( |
| 145 | "Plan revision %d accepted: %d phase(s), %d sub-step(s). The host renders it for the user and hands it to the executor — do not restate it.%s%s", |
| 146 | plan.Revision, phases, subSteps, approval, revised), nil |
| 147 | } |
| 148 |