declined

Look up the registry of declined and deferred decisions, so the same idea is never re-pitched: what was rejected, why, and under what condition to revisit. Read-only, and can also run the detector for

SKILL.md

GitHub

OBJECTIVE: Show the registry of declined/deferred decisions (what · why · revisit-if) BEFORE (re-)pitching an idea — the anti "step on the same rake twice" guard. READ-ONLY: it writes nothing into the registry (the nightly scan and the human do that).

CONTEXT:

  • The canonical registry (this hub): %USERPROFILE%\.claude\projects\<hub-project>\memory\declined-decisions.md (a copy exists in the laptop project — always read the hub one).
  • The nightly detector: $IMPORTS_ROOT/declined-scan/declined_scan.py (scheduled task "Declined-Decisions Nightly Scan", ~03:45) — it catches new refusals from fresh sessions and parks them on the AUTO-CAPTURED shelf.
  • The rule (cross-actor): rejected or deferred a proposal → record it in the registry. The canon for human assistants lives in the Bible as the "record a rejected decision" rule.

STEPS:

  1. Read declined-decisions.md (the hub path above). 1b. Dead-man check on the watchdog (added 2026-07-04; catches the class "the task was silently disabled", as happened 06-22→06-27): read $IMPORTS_ROOT/declined-scan/highwater.json → if updated is older than 2 days the nightly watchdog is NOT running → check (Get-ScheduledTask -TaskName 'Declined-Decisions Nightly Scan').State, enable it (Enable-ScheduledTask), and report in one line. Fresh → move on silently.
  2. If the operator asked about a SPECIFIC topic — grep the registry for it and show the matches (what was rejected · why · under what condition to revisit).
  3. Otherwise — a short summary: how many entries, the most recent additions, and the AUTO-CAPTURED shelf (waiting for promotion).
  4. (Optional, on "run the scan") execute python $IMPORTS_ROOT/declined-scan/declined_scan.py and show the newly caught refusals.
  5. If the idea being pitched right now is ALREADY in the registry — warn the operator explicitly ("this was declined on <date>, reason X, revisit if Y") before going further.

OUTPUT: a short list of relevant refusals, or a registry summary. It writes NOTHING (except step 4, where the scan writes for itself). Finish the reply to a non-technical operator with an "In plain words" recap.

RELATION (do not duplicate): the source registry = memory [[declined-decisions]]; the nightly scan = declined_scan.py; the rule for humans = the Bible entry on recording rejected decisions.

<!--kit-footer-->

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.