Awesome Reviewers

When a workflow stage can legitimately have missing data, treat the missing artifact as an explicit, validated state—not as an absent file/implicit default.

Apply this in two ways:

1) Declare optional inputs explicitly (so tooling can’t miss the leak)

• If downstream logic can use something only when it exists (e.g., a decision pack), declare it as an optional consume in the consuming stage.
• Ensure upstream validation (e.g., coverage/guards) verifies the dependency is actually wired and referenced.

2) On any “skip/jump” path, still write the required record as an explicit empty fact

• If control flow skips a questionnaire step, you must still create the artifact that records “None/empty” rather than omitting it.
• This prevents downstream stages from treating “missing file” as “unknown,” which is effectively a null.

Minimal pattern

• Use optional consumes for legitimately absent artifacts.
• Use an “empty inventory”/placeholder record for required facts.

Example (stage metadata intent)

# consuming stage
consumes:
  - artifact: decision-pack
    required: false   # optional, but explicitly declared
  - artifact: intent-statement
    required: true

<!-- skip path: still create the artifact -->
# source-inventory.md
sources:
  - type: none
    note: "Recorded during skip-ahead; questionnaire did not run."

Outcome: downstream stages never rely on silent nulls (missing artifacts, non-written placeholders, or prose-only fallbacks) and null/unknown states are either explicitly represented or validated away.

This site is open source. Improve this page.