@anicca self-improve: slop-detected

[Daisuke134]

Detected by self-improve loop (severity: warn)

Evidence: eval post-to-x scored 0.25 (<0.7) at 2026-06-04T15:49:38Z

Affected skill: anicca-x-useful

@anicca please investigate and propose a fix.

filed automatically by skills/self-improve

[Daisuke134]

Note: filed during P10 self-improve loop E2E verification (#335). The detection is genuine (reflects real live state), kept open for the loop to act on.

[Daisuke134]

👀 Anicca picked this up — tracking here.

stage	status
ack	✅ claimed
discuss	✅ responded

Round 1 — opinion updated from thread.

---

I've claimed this. A 0.25 on post-to-x means the output is barely usable—likely generic summary text with no hook or angle.

My audit plan: check the skill's prompt template and the last 10 published posts against the eval rubric. Leading hypothesis: anicca-x-useful is drafting posts without a mandatory content-judge gate, or it's pulling stale slots from workspace/hooks without freshness checks.

Proposed fix: (1) inject a content-judge pre-publish gate if bypassed, (2) rewrite the prompt to require a concrete insight or contrarian take, not bland summaries, (3) expire hooks older than 48h so we don't recycle decayed content.

Running the eval replay now. Will post the rubric breakdown shortly.