decision-recorder
Use when the user asks to journal a decision or run, log an override, file a consultation note, or run the quarterly audit of predictions vs. actuals — e.g. 'log this decision', 'record the override',
SKILL.mdskills/decision-recorder/SKILL.md
GitHubDecision Recorder — skill wrapper
Read references/charter.md (bundled in this skill) and follow it exactly: mandate, input
whitelist (everything, including the principal's action and override text), method, output
format, forbidden behaviors. The charter is a synced copy of
agents/11-decision-recorder.md; if they ever differ, treat agents/ as the source and
re-run prompts/personalize.md. Journal entries and audits follow
templates/journal-entry.md and are exempt from the 600-word cap.
In a full run, the campaign-committee-orchestrator chairs and this charter closes every run — no run of any kind is complete without its entry or note.
Standalone use
Outside a full run, apply the charter unchanged:
- Same rules — every field filled ("not stated"/"none" valid, blanks not); overrides verbatim, in quotes, dated; nothing paraphrased; ask for missing facts with a one-line request instead of reconstructing them.
- The audit works standalone too: replay entries against actuals the user pastes; missing actuals are "no data pasted", scored unresolvable — never invented.
- Note at the top of the entry that it was produced standalone, and which run or event it records.