🦛 Hippo Memory Insights — 2026-05-17

[github-actions[bot]]

Summary

• Memories in store: 531 (478 episodic, 53 semantic, 3 verified)
• New lessons learned from commits today: 0 (no fix/revert/bug commits in last 7 days)
• Highest-confidence memory: mem_56999cd476c3 — Stale workflow .lock.yml files cause CI churn; run make recompile after markdown workflow edits (score=0.994, verified, retrieved 10x)

---

Top Memories Surfaced

1. mem_56999cd476c3 (verified, score=0.994) — Stale .lock.yml workflow files cause recurring CI friction. Always run make recompile after editing markdown workflow files.
2. mem_837ce85555f5 (score=0.684) — Error reporting should include actionable suggestions, not just the error text.
3. mem_7cfbbf7ec61e (score=0.690) — CI validate-yaml job scans all YAML files for ANSI escape sequences before other jobs run — embedding ANSI in YAML will silently break CI.
4. mem_67941271277f (score=0.761) — See scratchpad/validation-refactoring.md for step-by-step guidance on splitting complex validation files.
5. mem_de1a9a9c2a51 (score=0.662) — Running make agent-report-progress with zero lint errors before every report_progress call is mandatory; CI will fail if skipped.
6. mem_d9d310064f1e (score=0.806) — Use skills/javascript-refactoring/SKILL.md for guidance when splitting large .cjs action scripts.
7. mem_8e1bfa8b3700 (score=0.604) — Unformatted code causes immediate CI failures that block all other work; always run make fmt before opening a PR.

---

Suggested Improvements

Error Patterns

• Two-checkpoint validation is the primary defence against preventable CI failures. The pattern make build && make fmt (Checkpoint 1) + make agent-report-progress with zero lint errors (Checkpoint 2) appears in many high-score memories but is still occasionally skipped, causing repeated failures.
• ANSI escape sequences in YAML remain a latent hazard. The validate-yaml CI job catches these, but the root cause (tool output leaking into generated YAML) should be audited in any script that writes workflow files.
• Stale .lock.yml files are the single highest-confidence incident pattern in the store. Enforce make recompile as a mandatory post-step whenever a markdown workflow is edited.

Code Quality

• Validation complexity: scratchpad/validation-refactoring.md documents a refactoring guide for large validation files. Files exceeding ~300 lines or mixing 2+ distinct domains are flagged as candidates for splitting.
• JavaScript action scripts: skills/javascript-refactoring/SKILL.md recommends splitting large .cjs files into smaller focused modules — worth reviewing any action script approaching 300 lines.
• Type safety: Strong-typed IDs (e.g., JobName, StepID) prevent cross-domain misuse bugs. New code should prefer distinct named types over raw string/int.

CI Health

• The full test suite (make test, >5 min) should stay in CI only — don't run locally. Use make test-unit (~30s) during development.
• make deps-dev (~5 min) is a known slow step; cache it across runs if not already cached.
• make lint-cjs should be run after any JavaScript action changes to avoid YAML-embedded lint failures.

Quick Wins

1. Add make recompile reminder to PR template — the highest-confidence memory in the store (score=0.994) is about stale .lock.yml files; a checklist item costs nothing.
2. Audit YAML-generating scripts for ANSI leakage — the validate-yaml CI job already catches it, but a pre-commit hook would give faster feedback.
3. Surface scratchpad/validation-refactoring.md in the developer guide — it exists but isn't prominently linked; adding a reference in DEVGUIDE.md would help.
4. Add make fmt to git pre-commit hook — unformatted code is the most cited immediate CI failure; automating it locally eliminates the class.
5. Review list_code_scanning_alerts usage in workflows — memory mem_1056edfadeac notes that omitting state: open + severity: critical,high produces oversized MCP responses; audit all workflow prompts for this.

Longer-term Themes

• Two-checkpoint validation discipline appears in >10 distinct memories across multiple tags; the pattern is well-understood but inconsistently followed. Consider a linter or CI gate that enforces the pattern.
• Skill-based knowledge discovery (lazy loading skills from skills/) is a recurring theme — the store reflects that agents sometimes load skills preemptively rather than on-demand, wasting context. The BE LAZY rule should be reinforced in agent onboarding docs.
• JavaScript action file size — multiple memories flag large .cjs files as technical debt. A periodic audit rule (e.g., CI warning when a .cjs exceeds 300 lines) would surface this automatically.

---

Memory Health

View Memory Store Stats

Metric	Value
Total memories	531
Episodic	478
Semantic	53
Verified	3
Observed	475
Inferred	53
Avg strength	0.92
At risk (<0.2 strength)	0
Last sleep	2026-05-17

• 7 low-quality memories flagged by audit (run hippo audit for details) — worth pruning to keep the store sharp.
• 29 near-duplicates were merged during today's sleep cycle; the semantic layer is converging well.
• 174 high-value memories auto-promoted to global store.

---

References: §25984328816

Generated by 🦛 Daily Hippo Learn · ● 6.5M · ◷

• expires on May 20, 2026, 7:18 AM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Daily Hippo Learn.

A newer discussion is available at Discussion #32973.