log-decision

Log a decision

When this skill fires

Activate when Shay states a decision has been made AS A COMPLETED ACT and needs to be recorded.

Trigger phrases (decision already made):

• "log this decision"
• "record this"
• "add to decisions log"
• "we decided X"
• "we agreed on Y — log it"
• "capture this decision"
• The slash command /log (renamed from /log-decision)

DO NOT activate for tentative, hedged, future-tense, or in-progress phrasing. Examples that look like decisions but aren't:

• "probably", "maybe", "likely" → tentative
• "we're thinking about", "leaning toward", "considering" → in-progress
• "I think we're going with X", "I want to go with X" → preference, not commitment
• "tomorrow we'll decide", "we'll probably pick", "next week we should pick" → future
• "we were thinking maybe X" → exploration

If the user appears to be thinking out loud or pre-deciding, do not activate. Instead, the chat layer can offer to capture the item as an open question (Q-NNN in the relevant document) but log-decision must not fire.

Also do not activate for:

• Action items, open questions, risks, or assumptions — those go to different files or sections.
• Decisions that contradict an existing active decision unless Shay also says "supersede D-NNN" or "reverse D-NNN".

This skill acts and reports — does not plan first

Logging a decision is a single append operation to a known file with a known format. Planning every entry would be theatrical friction. This skill:

1. Confirms a minimal set of fields
2. Appends the entry
3. Reports

If fields are ambiguous, asks before writing. Otherwise proceeds.

Required reads before logging

1. The target project's decisions-log.md — to read the last D-NNN and increment correctly
2. D:\Claude-Cowork\00_Context\working-preferences.md — to confirm format conventions

Fields required for every entry

Mandatory:

• Title — short, present-tense decision phrasing (e.g., "Use PostgreSQL over MongoDB for primary store", "Defer client portal to phase 2")
• Decision — one or two sentences: what was decided
• Rationale — why: reasoning, constraints, evidence

Recommended (ask if not obvious from context):

• Alternatives considered — at minimum one rejected alternative with reason
• Decided by — name(s) / role(s). Default to Shay if not specified.

Auto-populated:

• ID — next D-NNN after reading the last entry
• Date — today's date YYYY-MM-DD
• Status — "active" unless this is superseding/reversing an existing decision
• Implications / follow-ups — only if Shay mentions them; otherwise leave blank with a placeholder for future updates
• References — workspace paths if Shay cites source files; otherwise leave blank

Entry format (mandatory)

• D-NNN —

• Date: YYYY-MM-DD

• Status: active | superseded by D-YYY | reversed on YYYY-MM-DD

• Decided by: <name(s) / role(s)>

• Decision:

• Rationale:

• Alternatives considered:

  • <alt 1> —
  • <alt 2> —

• Implications / follow-ups: <or "TBD">

• References: <or "—">

Placement in the file

Insert directly below the "How to use" section header, ABOVE any existing entries. Reverse chronological order is enforced — newest at top, always.

If the file's existing entries are not in reverse chronological order, do NOT silently reorder them. Report the inconsistency and ask whether to fix it as a separate operation.

Superseding / reversing existing decisions

If Shay says "supersede D-NNN" or "reverse D-NNN":

1. Create the new entry as normal with next D-ID.
2. Update the old entry's Status field:
   • superseded by D-YYY on YYYY-MM-DD — keep entry, mark relationship
   • reversed on YYYY-MM-DD by D-YYY — keep entry, mark relationship
3. Do NOT delete the old entry. Audit trail is sacred.
4. In the new entry's Rationale, reference the superseded/reversed ID and explain what changed.

Voice check

Apply the AVOID-list to the entry text. If a forbidden word appears in Shay's verbatim quote, preserve it (it's a quote) but flag. If it appears in your generated phrasing (rationale, alternatives), rewrite before saving.

After logging — report

• ID assigned (D-NNN)
• Full file path
• One-line confirmation of what was logged
• Any inconsistencies noticed (e.g., out-of-order entries, missing fields in nearby entries)
• For Command Center (only if Shay mentioned follow-ups): the implications/follow-up items as candidate tasks

Failure modes

• decisions-log.md doesn't exist for this project: stop. This means the project wasn't initialized through new-project-setup. Ask whether to create the file with proper structure first, or whether the wrong project is being targeted.
• Last D-NNN is malformed or unclear: stop. Report and ask Shay how to proceed (manually specify the ID, or fix the file first).
• Two concurrent decisions in one request: log them as separate entries (D-NNN and D-NNN+1), not as one merged entry. One decision per entry, always.
• Shay describes something that isn't actually a decision (e.g., an open question framed as a decision): stop. Reflect back what was heard, ask for clarification. Don't pollute the decisions log with non-decisions.

What this skill is NOT for

• Logging action items — those belong in meeting summaries (Section 2) or in the Command Center.
• Logging assumptions — those go in the relevant document's Assumptions section with A-NN IDs.
• Logging risks — those go in the relevant document's Risks section with RSK-NN IDs.
• Logging open questions — those go in the relevant document's Open Questions section with Q-NN IDs.
• Bulk-importing decisions from elsewhere — that's a one-time migration, not this skill's job.