Issues

• checkpoint-stop-hook: measure '## Last session' cap (20 entries / 200 chars) before demanding an update — recurred to 308 entries 9 days after manual diet

  caty-ai/caty-agent-harness#298 · Filter by author shojikumaru opened on Sep 11, 2026

• lib-state-fold: pin (and consider narrowing) the id-prefix strip after the fold date in normalize_state_candidate (follow-up to #279 / PR #292)

  caty-ai/caty-agent-harness#294 · Filter by author shojikumaru opened on Sep 7, 2026

• distill-audit: a newline in a candidate filename forges an extra TSV row and pulls an unrelated file into the distiller prompt (pre-existing; follow-up to #151)

  caty-ai/caty-agent-harness#293 · Filter by author shojikumaru opened on Sep 7, 2026

• learning-path probe: the 'distiller cron' conjunct does not probe each adapter's real scheduling mechanism (follow-up to #57 / PR #286)

  caty-ai/caty-agent-harness#288 · Filter by author shojikumaru opened on Sep 6, 2026

• checkpoint hooks: session_id sanitizer must not abort on invalid UTF-8 under pipefail (all three adapters)

  caty-ai/caty-agent-harness#285 · Filter by author shojikumaru opened on Sep 6, 2026

• tests: pin the __pycache__/ and *.pyc ignore rules (follow-up to #233 / PR #281)

  caty-ai/caty-agent-harness#284 · Filter by author shojikumaru opened on Sep 6, 2026

• learning loop: recurrence watch after promotion (same theme seen again within N days goes to Open failures)

  component:stateenhancement
  caty-ai/caty-agent-harness#280 · Filter by author shojikumaru opened on Sep 6, 2026

• apply-promotions: a skill candidate with promote: not-yet is staged anyway (EV-007c observation)

  ball:human
  caty-ai/caty-agent-harness#278 · Filter by author shojikumaru opened on Sep 6, 2026
  3

• apply-promotions: a re-stated theme supersedes for rules but parks as supersedes-ambiguous for facts/skills — invalidated-by chains grow by wording (EV-007c observation)

  caty-ai/caty-agent-harness#277 · Filter by author shojikumaru opened on Sep 6, 2026

• follow-ups from the #263 review: rollback-vs-config gate, first-prefix session bind, session charset fallback, run-k trust note, doc NITs

  caty-ai/caty-agent-harness#268 · Filter by author shojikumaru opened on Sep 2, 2026

• EV-007c: learning-loop measurement on Sonnet and Opus — per-model difficulty-calibration pilot first, then the EV-007b protocol per model

  ball:human
  caty-ai/caty-agent-harness#265 · Filter by author shojikumaru opened on Sep 2, 2026
  11

• rubric: templates and bootstrap instruct filling a rubric that nothing scores (2 adopted nightly findings)

  component:task-runnerdocumentationseverity:low
  caty-ai/caty-agent-harness#253 · Filter by author shojikumaru opened on Aug 31, 2026
  1

• install: flush-intake consumer is never scheduled — flush-*.md accumulate silently (adopted nightly finding)

  component:adapterscomponent:installerintegration:claude-codeseverity:medium
  caty-ai/caty-agent-harness#252 · Filter by author shojikumaru opened on Aug 31, 2026

• adapter hooks perf: per-turn full-tree find scans, synchronous 120s PreCompact flush, per-file stat/wc in distill-audit (3 adopted nightly findings)

  component:adaptersintegration:claude-codeintegration:codexintegration:kimiintegration:openclawseverity:low
  caty-ai/caty-agent-harness#251 · Filter by author shojikumaru opened on Aug 31, 2026

• task-runner perf batch: per-tick full artifact re-scans + state-fold subprocess overhead (7 adopted nightly findings)

  component:statecomponent:task-runnerseverity:low
  caty-ai/caty-agent-harness#250 · Filter by author shojikumaru opened on Aug 31, 2026

• untrusted input → shell: distill-audit awk system() sink + stale-claim-watchdog job_id validation (2 adopted nightly findings)

  component:adaptersintegration:hermesintegration:openclawsecurityseverity:medium
  caty-ai/caty-agent-harness#244 · Filter by author shojikumaru opened on Aug 31, 2026
  3

• sentinel cleanup: unify regime-state construction and null-cache replay semantics (deferred from the #218 panel)

  caty-ai/caty-agent-harness#237 · Filter by author shojikumaru opened on Aug 30, 2026
  1

• pr-size: a body-only edited event voids the size-exempt label and re-fails the check post-merge (head unchanged)

  caty-ai/caty-agent-harness#227 · Filter by author shojikumaru opened on Aug 29, 2026
  1

• raw-review follow-ups: dead truncation branch, pause indeterminate label, lock-busy notify routing, conf template location

  caty-ai/caty-agent-harness#196 · Filter by author shojikumaru opened on Aug 26, 2026

• OpenClaw distill-audit route: port the '## Last session' fold (parity with claude-code/hermes)

  caty-ai/caty-agent-harness#173 · Filter by author shojikumaru opened on Aug 24, 2026

• task-runner: fail loud on zero-byte model spawns instead of burning the full attempt budget as generic timeouts

  caty-ai/caty-agent-harness#162 · Filter by author shojikumaru opened on Aug 23, 2026

• feat(design): adaptive activation — overflow sentinel fires the task-runner from measured per-turn context re-injection

  external-input
  caty-ai/caty-agent-harness#159 · Filter by author shojikumaru opened on Aug 22, 2026
  21

• learning loop: monthly stocktake of the durable tier (consolidate/demote proposals -> human decision -> compaction) + implement the specified rollback path

  caty-ai/caty-agent-harness#149 · Filter by author shojikumaru opened on Aug 21, 2026
  2

• EV-006 rig / lane 2: local model on Ollama @ VPS (qwen3:8b) — CLAUDE_BIN shim, sealed-diff prereg, scaled-down run

  caty-ai/caty-agent-harness#143 · Filter by author shojikumaru opened on Aug 21, 2026
  1

• tracking: multi-model benchmark lanes on the sealed EV-006 rig (Luna, Ollama/VPS, …)

  caty-ai/caty-agent-harness#129 · Filter by author shojikumaru opened on Aug 19, 2026
  7