Operationalize Recurrent Activities Through Processes
Status: open/published
Summary
Section titled “Summary”- Any reoccurring workflow becomes a defined process: captured, executable, repeatable. This principle owns the capture decision (when repetition crosses a practical threshold); maturation of the captured process is P-0001-prefer-deterministic-over-llm.
- Conversational and brainstorming work may originate a process, and is a step inside one — not an alternative to one.
- Skills are the initial operationalization layer; without activity-to-process as the default, schema and determinism have nothing to optimize. Refer to Workflow Evolution principle.
Statement
Section titled “Statement”The default form for any recurring activity in sdlc is a defined process: captured, executable, repeatable. Conversational and brainstorming work isn’t an alternative to processes; it’s a step inside one. The system’s posture: if an activity is a reoccurring workflow — it will be performed multiple times over more than a single work session — it is worth capturing in an enduring workflow expression. That recurrence test is the capture threshold — work that stays too rare or too fuzzy to recur isn’t forced into a process. How far a captured process then hardens is governed by P-0001-prefer-deterministic-over-llm.
Ad-hoc work doesn’t compound. Each repetition starts from a blank slate; lessons live in someone’s head or not at all; quality varies with attention. Putting the activity into a process (even a thin one) lets the next run reuse the substance of the last, lets the form get critiqued and improved as a first-class artifact, and lets the activity scale beyond the one person who originally did it.
This principle is the foundational rule the others follow from. Without “activity to process” as the default, P-0001-prefer-deterministic-over-llm has nothing to optimize; P-0005-schema-over-prose has nothing to schema-ify.
How it applies
Section titled “How it applies”- New work that crosses the capture threshold (a reoccurring workflow — it recurs across more than a single work session) gets a process: at minimum a skill or checklist, even before deterministic extraction.
- Conversational steps (brainstorming, requirement-gathering, retrospective) are named, structured steps inside a process, not “the part that’s free-form because LLMs are good at conversation.” Even free-form conversation has a defined beginning, end, and output.
- Tasks themselves are an instance of this principle: they’re the operationalization of “the work to be done.” See C-0001 Executable Task for the realization.
Examples
Section titled “Examples”/sdlc:task-define: the recurring ‘create a new task from a fuzzy idea’ activity, captured as a defined process./sdlc:backlog-triage: the recurring ‘review backlog and promote’ activity, with named steps and an output shape./sdlc:pr-respond: PR-comment handling as a structured workflow rather than ad-hoc reply.- The skill ADR’s required shape (Status, Trigger, Steps, Outputs, Post-execution assessment, Failure modes): the structure every operationalized process takes.
Implications
Section titled “Implications”- The skill surface is typically the first operationalization layer. Every skill is
one process; the skill ADR’s required shape (Status, Trigger, Steps,
Outputs, Post-execution assessment, Failure modes) is the structure
a process takes. See
docs/planning/decisions/D-TQHZ-skill.md. - The natural-progression model in P-0001-prefer-deterministic-over-llm is what processes do over time as they mature: ad-hoc, then process, then process with scripts, then deterministic workflow with judgment ports.
- Post-execution assessment (per the skill ADR) is the mechanism by which processes improve themselves. Without retrospection baked in, operationalization is incomplete.