Log Decision

Skill: log-decision

Capture the why behind a choice so future-you (or a collaborator, or a reviewer) can reconstruct the reasoning, not just the result. Decisions are recorded as tracked log entries in the ledger — provenance for judgment, alongside the provenance for computation. You delegate the save to the datalad doer.

When to use

• A non-obvious methodological or design choice was made (a model, an exclusion criterion, a tool,

dropping/keeping a comparison) and its rationale should be preserved.

• Do NOT use to record a computation (analyze/run-comparison already provenances that) or to make

a formal commitment (govern/preregister / govern/obligations).

Steps

1. Capture the decision — get: the decision (what was chosen), the rationale (why), the

alternatives considered and why they were not chosen, and the scope (which branch/comparison/ product it affects, if any).

1. Append to the log — one entry to project.yaml log[]:

{ ts, op: log-decision, stage: manage, note: "DECISION: — WHY: — ALT: ", branch: }. Keep the note self-contained so the reasoning is legible from the log alone.

1. Save — delegate to the datalad doer: "save: datalad save -m 'log-decision: '."
2. Report — confirm the decision was recorded and on which branch.

Constraints

• Record the decision and its rationale and alternatives — a decision without its "why" is not

worth logging; do not reduce it to a bare statement.

• Record honestly: capture the actual reasoning, including trade-offs, not a post-hoc justification.
• log: is append-only — a reversed decision is a new log-decision entry referencing the prior

one, never an edit of it. Keep the ledger schema-valid; delegate the save to the datalad doer.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: neuroanalytics
• Source: neuroanalytics/data-science-harness
• License: MIT

Install and usage instructions live in the source repository linked above.