🦛 Hippo Memory Insights — 2026-05-12

[github-actions[bot]]

Summary

• Memories in store: 527 (500 active + 27 newly merged semantic)
• New lessons learned from commits today: 0 new (1 duplicate skipped — awfVersionAtLeast helper extraction already known)
• Highest-confidence memory: Stale .lock.yml files causing CI churn (score=1.007, verified incident)

---

Top Memories Surfaced

1. mem_b4a7d0d8f2dc (score=1.007, verified) — Stale workflow .lock.yml files cause CI churn and friction; always run make recompile after editing markdown workflow files.
2. mem_7cf9de66a832 (score=0.594) — Unformatted code has caused 5 CI failures in a single day; make fmt is mandatory before every commit.
3. mem_e573e8db1995 (score=0.699) — validate-yaml job in CI scans all YAML for ANSI escape sequences before any other jobs run; never embed terminal color codes in YAML.
4. mem_09048a4ad4b3 (score=0.671) — CI will fail if make recompile is skipped after markdown edits — this is automatic and non-negotiable.
5. mem_7d523a42edbf (score=0.816) — skills/javascript-refactoring/SKILL.md is the canonical guide for splitting JS into .cjs files; consult it before any JS restructuring.
6. mem_c4e60f5a58df (score=0.772) — scratchpad/validation-refactoring.md has a step-by-step guide for the ongoing validation-layer refactoring effort.
7. mem_5c3420591939 (score=0.183) — Architecture uses runtime file copying (not embedded scripts) for JS/shell script distribution.

---

Suggested Improvements

Error Patterns

• Lock file drift is the top recurring incident (score=1.007, verified, seen multiple times). Automate make recompile as a pre-commit hook or add a CI check that fails fast if compiled outputs are stale relative to their markdown sources. This alone would prevent the most common class of CI failures.
• Formatting failures are the second most common cause of wasted agent sessions. A pre-commit hook running make fmt would eliminate these entirely with zero developer overhead.

Code Quality

• scratchpad/validation-refactoring.md describes an in-progress refactoring of the validation layer (retrieved 2×, high confidence). This work should be tracked to completion — the scratchpad file suggests it is partially done.
• Type-safety patterns (mem_bffac2406682) are well-documented but retrieved infrequently; consider adding a linter rule or review checklist item to enforce them for new JobName/StepID style typed strings.
• File-size discipline: the memory store consistently recommends files <300 lines and splitting at domain boundaries. A make lint check on file length could enforce this automatically.

CI Health

• validate-yaml ANSI check — the CI job position (runs before everything else) is correct. No changes needed, but contributors should be reminded in CONTRIBUTING.md that ANSI codes in YAML are a hard failure.
• make test (>5 min) — the team already knows to let CI handle the full suite. make test-unit (~3 min) is the right local gate. Document this more prominently in DEVGUIDE.md so new contributors don't block on slow local runs.
• 7 low-quality memories were flagged by hippo audit — run hippo audit and prune them to keep recall precision high.

Quick Wins

1. Pre-commit hook: make fmt && make recompile — eliminates the top two CI failure modes.
2. Update DEVGUIDE.md — add explicit note that make test-unit is the local gate; make test is CI-only.
3. Prune 7 low-quality memories — run mcpscripts hippo --args "audit" and remove stale/contradictory entries to sharpen recall.
4. Track validation-refactoring to completion — the scratchpad guide exists; file an issue or PR to close out the remaining validation-layer splits.
5. Add CONTRIBUTING.md note about ANSI codes in YAML being a hard CI blocker.

Longer-term Themes

• Compiled-output drift (.lock.yml, recompiled JS) appears repeatedly across multiple sessions. A systemic fix — either a CI diff check or a build step that regenerates outputs and fails on diff — would eliminate this permanently rather than relying on developer discipline.
• Validation architecture is a recurring refactoring theme. The memory store references both scratchpad/validation-refactoring.md and the skills/developer skill multiple times, suggesting the validation layer is an area of ongoing churn worth dedicated effort.
• Skill/memory hygiene: 52 new semantic memories were created this cycle from episodic merges, and 25 cross-layer duplicates were pruned. The store is healthy but growing fast — periodic hippo audit + prune sessions (monthly?) will keep quality high.

---

Memory Health

Memory store statistics

Metric	Value
Total memories	527
Active	500
Decayed (removed)	0
Episodic merged → semantic	125
New semantic created	52
Cross-layer duplicates pruned	25
Low-quality flagged	7
Embeddings	527/527 (100%)

Memories are mostly fresh (<7 days), high average strength, diverse topics — store is in steady, healthy operation.

Action needed: Run mcpscripts hippo --args "audit" to review the 7 flagged low-quality memories and decide which to prune.

---

References: §25719451445

Generated by Daily Hippo Learn · ● 5.9M · ◷

• expires on May 15, 2026, 7:21 AM UTC