Plan Workflow
Use this skill to turn a vague or execution-sensitive request into a compact
user mandate. The mandate feeds $tcx-workflow for immediate runs and
$tcx-automate for recurring runs. It is conversation context, not a
server-generated plan, semantic lane, selected team, role ceiling, or DAG. Head
Manager still interprets the live request, begins the lightweight analysis run,
and dynamically chooses or revises exact fixed roles from accepted evidence.
Procedure
- Restate the user's objective, subject scope, time or as-of horizon, requested
output, and whether the work is immediate or recurring.
- Extract binding constraints: symbols or universe when supplied, Strategy or
account scope when explicitly selected, allowed actions, blocked actions,
approval posture, evidence requirements, stop conditions, and schedule when
recurring.
- Ask only for missing fields whose answer would materially change the result,
suitability, action authority, automation arming, or stop condition. Leave
analytical decomposition and role choice to
$tcx-workflow.
- Preserve explicit prohibitions in the user's own terms. Do not translate
them into a keyword classification, fixed role list, or default team.
- Write reader-facing questions, summaries, and the
Workflow Mandate in the user's language from the original request unless the user explicitly asks for another language.
- Produce a compact
User Mandate with the fields below, then hand off:
- immediate investment work:
READY_FOR_DYNAMIC_ANALYSIS
- recurring workflow:
READY_FOR_AUTOMATION_PREFLIGHT
- missing material fields:
NEEDS_CLARIFICATION
- server/build request:
SERVER_OR_BUILD_HANDOFF
- unsafe or unsupported request:
BLOCKED_BY_SCOPE
Question Discipline
- Ask at most three focused questions at a time.
- Do not assign a default analytical category, stage sequence, or analyst team to
a broad request. Ask only when unresolved scope would materially change the
requested outcome; otherwise let Head Manager begin with the smallest useful
question.
- Do not block general research on broker, account, or profile details that are
irrelevant to the requested output.
- Require explicit answers for order drafting, paper execution, live execution, account/broker scope, pre-approved mandates, and recurring schedules.
- If automation is requested, require schedule, blocker policy, and expiry or rearm condition before
READY_FOR_AUTOMATION_PREFLIGHT.
Mandate Shape
Return this shape in prose or compact YAML:
user_mandate:
status: READY_FOR_DYNAMIC_ANALYSIS
objective: ""
requested_output: ""
subject_scope: []
as_of_or_horizon: ""
recurring: false
schedule: ""
explicit_strategy_or_account_scope: ""
investor_context_needed: []
broker_scope: ""
allowed_actions: []
blocked_actions: []
approval_model: ""
evidence_requirements: []
preflight_checks: []
stop_conditions: []
unresolved_questions: []
Keep field names, file paths, symbols, tickers, source names, and quoted source
text in their natural/original form. Do not add role candidates, stages, a lane,
or a server workflow id. begin_analysis_run owns the durable request hash and
sealed Brain, Strategy, and Investor Context provenance; it does not persist
this raw mandate.
Hard Stops
- Do not dispatch subagents, choose a role team, produce investment analysis,
approve orders, execute orders, or register automations.
- Do not create or imply a server-side semantic plan, lane, task queue, or DAG.
- Do not turn natural language into approval or execution authority.
- Do not ask for raw secrets or direct broker/API access.
- Do not widen a user's mandate while resolving ambiguity.
1---2name: tcx-plan3description: Clarify an ambiguous TradingCodex mandate before immediate or recurring work by asking focused questions and preserving scope, constraints, action boundaries, approval posture, and stop conditions. Use when the user explicitly asks to plan, scope, or stress-test investment work, or when a missing choice in a schedule or automation would materially change the result or authority boundary. Clear recurring requests should use tcx-automate directly.4---56# Plan Workflow78Use this skill to turn a vague or execution-sensitive request into a compact9user mandate. The mandate feeds `$tcx-workflow` for immediate runs and10`$tcx-automate` for recurring runs. It is conversation context, not a11server-generated plan, semantic lane, selected team, role ceiling, or DAG. Head12Manager still interprets the live request, begins the lightweight analysis run,13and dynamically chooses or revises exact fixed roles from accepted evidence.1415## Procedure16171. Restate the user's objective, subject scope, time or as-of horizon, requested18 output, and whether the work is immediate or recurring.192. Extract binding constraints: symbols or universe when supplied, Strategy or20 account scope when explicitly selected, allowed actions, blocked actions,21 approval posture, evidence requirements, stop conditions, and schedule when22 recurring.233. Ask only for missing fields whose answer would materially change the result,24 suitability, action authority, automation arming, or stop condition. Leave25 analytical decomposition and role choice to `$tcx-workflow`.264. Preserve explicit prohibitions in the user's own terms. Do not translate27 them into a keyword classification, fixed role list, or default team.285. Write reader-facing questions, summaries, and the `Workflow Mandate` in the user's language from the original request unless the user explicitly asks for another language.296. Produce a compact `User Mandate` with the fields below, then hand off:30 - immediate investment work: `READY_FOR_DYNAMIC_ANALYSIS`31 - recurring workflow: `READY_FOR_AUTOMATION_PREFLIGHT`32 - missing material fields: `NEEDS_CLARIFICATION`33 - server/build request: `SERVER_OR_BUILD_HANDOFF`34 - unsafe or unsupported request: `BLOCKED_BY_SCOPE`3536## Question Discipline3738- Ask at most three focused questions at a time.39- Do not assign a default analytical category, stage sequence, or analyst team to40 a broad request. Ask only when unresolved scope would materially change the41 requested outcome; otherwise let Head Manager begin with the smallest useful42 question.43- Do not block general research on broker, account, or profile details that are44 irrelevant to the requested output.45- Require explicit answers for order drafting, paper execution, live execution, account/broker scope, pre-approved mandates, and recurring schedules.46- If automation is requested, require schedule, blocker policy, and expiry or rearm condition before `READY_FOR_AUTOMATION_PREFLIGHT`.4748## Mandate Shape4950Return this shape in prose or compact YAML:5152```yaml53user_mandate:54 status: READY_FOR_DYNAMIC_ANALYSIS55 objective: ""56 requested_output: ""57 subject_scope: []58 as_of_or_horizon: ""59 recurring: false60 schedule: ""61 explicit_strategy_or_account_scope: ""62 investor_context_needed: []63 broker_scope: ""64 allowed_actions: []65 blocked_actions: []66 approval_model: ""67 evidence_requirements: []68 preflight_checks: []69 stop_conditions: []70 unresolved_questions: []71```7273Keep field names, file paths, symbols, tickers, source names, and quoted source74text in their natural/original form. Do not add role candidates, stages, a lane,75or a server workflow id. `begin_analysis_run` owns the durable request hash and76sealed Brain, Strategy, and Investor Context provenance; it does not persist77this raw mandate.7879## Hard Stops8081- Do not dispatch subagents, choose a role team, produce investment analysis,82 approve orders, execute orders, or register automations.83- Do not create or imply a server-side semantic plan, lane, task queue, or DAG.84- Do not turn natural language into approval or execution authority.85- Do not ask for raw secrets or direct broker/API access.86- Do not widen a user's mandate while resolving ambiguity.
monarchjuno