2026-06-05 wiki-lint-daily
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 2/2 | Excellent — 14-check scan, exact file counts, specific broken-wikilink categories. Concrete artifact paths. |
| FOLLOW-UP | 2/2 | ”Recommendations for Next Session” with 4 actionable items ranked by priority. |
| WIKI-INTEGRATION | 2/2 | Session file created, cites SCHEMA.md, index.md trends vs prior run. |
| SIGNAL-TO-NOISE | 1/2 | Good structure but ~15 lines of prose for what could be 6. Slight padding in trend commentary. |
| SCOPE-DISCIPLINE | 1/2 | Stayed within budget but did not auto-fix anything — deferred to “report findings before fixing” convention (valid, but should note that explicitly). |
Error: RuntimeError: HTTP 503: Loading model — the inference server (.106) was busy/unavailable. This is the same failure mode seen in agent-tool-watch, hermes-queue-infra, mercury-skills-ingest, hermes-tools-audit today (all at ~13:00–16:00 UTC).
Prompt delta suggestion: Add a retry instruction: “If inference server returns 503, wait up to 60s and retry once. If still failing, log BLOCKED with timestamp and skip.” — 4 jobs failed today from the same root cause and could have auto-retried.
2026-06-05 Smart Groceries
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 2/2 | Table of blockers with exact status, file paths (.git-credentials), MR !1 reference. |
| FOLLOW-UP | 2/2 | Clear “Requires pvs to” action list at bottom. Previous session links provided. |
| WIKI-INTEGRATION | 2/2 | Session file created, wikilinks to prior sessions and root cause entry. |
| SIGNAL-TO-NOISE | 2/2 | Tight — no meta-narration, just facts and blockers. |
| SCOPE-DISCIPLINE | 1/2 | Stayed within budget (~2 min). But this is a recurring “nothing changed” pattern for 10+ days — should escalate to [SILENT] or add a flag like no_change_for: 10d. |
Note: This job produces value by monitoring, but the same output has been generated daily since May 25. Recommend either reducing frequency (daily → every 3 days when blocked) or adding a consecutive_no_change counter to auto-suppress delivery.
2026-06-05 daily-bill-scan
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 2/2 | Concrete: file paths, version numbers (PyMuPDF v1.27.2.3), OCR results list with filenames and vendors. |
| FOLLOW-UP | 2/2 | ”Issues / Questions” section with two specific decision points for pvs (PowerCo test data, Superloop dummy account). |
| WIKI-INTEGRATION | 1/2 | Session file created but no wikilinks to existing bill-scan wiki pages. Project index should be [[projects/bill-scan/index]] if it exists — check actual project slug. |
| SIGNAL-TO-NOISE | 2/2 | Very tight — fragments only, no meta-narration. |
| SCOPE-DISCIPLINE | 2/2 | Completed in ~5 min (well within 15-min budget). Fixed a real bug (shebang + sys.path hack) as part of the work. |
Prompt delta suggestion: Add “cite or create wiki page before starting” instruction: always link to [[projects/daily-bill-scan/index]] in the session file for traceability.
2026-06-05 archive-inactive-sessions
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 0/2 | FAILED — RuntimeError: Connection error. No output. |
| FOLLOW-UP | 0/2 | Failed. |
| WIKI-INTEGRATION | 0/2 | No session file created. |
| SIGNAL-TO-NOISE | N/A | N/A |
| SCOPE-DISCIPLINE | N/A | N/A |
Error: RuntimeError: Connection error. — generic network or API failure. Distinct from the HTTP 503 cluster (this ran at 05:08 UTC, before the inference server load spike). Likely a one-off transient failure.
Prompt delta suggestion: Add retry-on-connection-error with exponential backoff (2s, 4s, 8s max 3 retries). Also add on_fail: create_session_log so even failed runs produce a session file documenting what happened.
2026-06-05 agent-tool-watch
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 0/2 | FAILED — HTTP 503: Loading model. Same cluster as wiki-lint-daily. |
| FOLLOW-UP | 0/2 | Failed. |
| WIKI-INTEGRATION | 0/2 | No session file. |
| SIGNAL-TO-NOISE | N/A | N/A |
| SCOPE-DISCIPLINE | N/A | N/A |
Error: Same HTTP 503 inference server issue at ~16:10 UTC. This job is particularly sensitive because it likely queries the model endpoint as its primary task — no fallback exists.
2026-06-05 asx-trading
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 2/2 | Excellent — file paths (fee_gate.py, test_fee_gate.py), test count (24 passing), capital tier table, gate checks. |
| FOLLOW-UP | 2/2 | Clear next step: “Dividend layer spec (Deliverable #3)“. Status field marked done. |
| WIKI-INTEGRATION | 1/2 | Session file created with frontmatter and wikilinks to prior sessions. Missing link to parent project index ([[projects/asx-trading/index]]). |
| SIGNAL-TO-NOISE | 2/2 | Tight, structured, no filler. Tables are dense and informative. |
| SCOPE-DISCIPLINE | 2/2 | Delivered one complete deliverable (Phase 3 #2). Focused work within budget. |
Prompt delta suggestion: Add explicit instruction to link parent project index: “Always include [[projects/<slug>/index]] in session frontmatter.”
2026-06-05 telemetry-daily-summary
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Output written to /opt/data/wiki/hermes/telemetry-summary/2026-06-05.md (concrete path). Stats are numeric but brief. |
| FOLLOW-UP | 1/2 | No explicit next-step — this is a monitoring job so “no anomalies” is the natural result. Acceptable. |
| WIKI-INTEGRATION | 1/2 | Writes to wiki path but no wikilinks. Output directory not linked from index. |
| SIGNAL-TO-NOISE | 2/2 | Terse — just facts, 4 lines of summary. |
| SCOPE-DISCIPLINE | 2/2 | Completed quickly, within budget. |
2026-06-05 telemetry-12h-spotcheck
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | File path provided, stats table present. Very brief output (single tick). |
| FOLLOW-UP | 1/2 | Same as daily — monitoring job, no action needed when clean. |
| WIKI-INTEGRATION | 1/2 | Writes to wiki path but minimal integration. |
| SIGNAL-TO-NOISE | 2/2 | Tight output, no padding. |
| SCOPE-DISCIPLINE | 2/2 | Fast completion. |
2026-06-05 hermes-self-improve
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Identified CrashLoopBackOff pods by name, described error types. But the session ended abruptly mid-thought (“Let me load the relevant skill…”). |
| FOLLOW-UP | 0/2 | Truncated — no next-step documented, no kanban update. Session left in incomplete state. |
| WIKI-INTEGRATION | 0/2 | No session file, no wiki links. |
| SIGNAL-TO-NOISE | 1/2 | Compact but the “Now I have a clear picture…” line is unnecessary meta-narration. |
| SCOPE-DISCIPLINE | 0/2 | Appears to have timed out or been interrupted mid-session — did not complete any kanban protocol steps (show → heartbeat → comment). |
Issue: Session was cut off. The agent started identifying CrashLoopBackOff issues but never completed the kanban protocol. This suggests either a timeout, tool failure during kubectl calls, or the agent getting stuck. Review logs for this run to determine cause.
2026-06-05 hermes-asx-trading
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Good error analysis: identified kanban tool bug with specific symptom (active board not propagating). Mentioned config file location. |
| FOLLOW-UP | 1/2 | Blocked but documented the issue and listed 3 projects in queue. Could have been tighter. |
| WIKI-INTEGRATION | 0/2 | No session file created despite ~40 lines of output. |
| SIGNAL-TO-NOISE | 1/2 | The error message is itself quite verbose (nested paragraph structure) — feels like tool output leaked into the response. |
| SCOPE-DISCIPLINE | 0/2 | Blocked immediately, did not attempt workaround. Acceptable for a tool bug, but should have logged to session file before failing. |
Error: Kanban tool caching/state bug — active board switch not propagating to subsequent list calls.
Prompt delta suggestion: Add “If blocked by tool bug, write a one-line session file first, then report” instruction.
2026-06-05 hermes-smart-groceries
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Repeated the blocker table (GitLab auth + Imperva). Same data as parent “Smart Groceries” cron job. File path references present. |
| FOLLOW-UP | 0/2 | Just repeats blockers — no new analysis, no alternative work attempted. |
| WIKI-INTEGRATION | 1/2 | Session file exists but identical content to the parallel Smart Groceries run above. |
| SIGNAL-TO-NOISE | 2/2 | Very terse — just facts. |
| SCOPE-DISCIPLINE | 2/2 | Completed quickly. Stayed within budget. |
Issue: Redundant with “Smart Groceries” job — both checked the same project, found the same blockers, produced near-identical output. These should be de-duplicated or one disabled.
2026-06-05 hermes-voice-pipeline
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Identified blocker (dispatcher profile collision) with error message and root cause analysis (reserved hermes binary name). Task IDs cited. |
| FOLLOW-UP | 0/2 | Same “pvs must either reassign or rename” — identical to previous runs since May 30. No alternative action attempted. |
| WIKI-INTEGRATION | 0/2 | Mentions kanban comments added but no session file visible in project path. |
| SIGNAL-TO-NOISE | 2/2 | Tight summary, table format effective. |
| SCOPE-DISCIPLINE | 1/2 | Completed quickly but has been running ~548 attempts since May 30 with identical failure. This job should be disabled until the dispatcher profile issue is fixed — it’s burning tokens on a known blocker. |
2026-06-05 hermes-synthetic-avatar
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 2/2 | Excellent research: ranked models with specific metrics (SoulX-FlashHead 96 FPS, Apache 2.0 license). Kanban task ID created (t_b3774ddc). |
| FOLLOW-UP | 1/2 | Completed research phase but got blocked on kanban task assignment. Clear next step implied (implementation phase) but not explicitly stated as a follow-up. |
| WIKI-INTEGRATION | 1/2 | Kanban entry created, but no session file in wiki/projects/synthetic-avatar/sessions/. Research should be linked to [[projects/synthetic-avatar/research]]. |
| SIGNAL-TO-NOISE | 2/2 | Tight research table, no fluff. |
| SCOPE-DISCIPLINE | 1/2 | Did one complete phase (research). Blocked after attempting kanban assignment but did not escalate cleanly — stuck in profile collision loop. |
2026-06-05 hermes-queue-infra, mercury-skills-ingest, hermes-tools-audit
| Rubric | Score | Note |
|---|---|---|
| ALL | 0/2 | All three failed with HTTP 503: Loading model — same inference server cluster failure. No output, no session files created. |
Cluster pattern: At least 4 jobs (wiki-lint-daily ~13:18, agent-tool-watch ~16:10, hermes-queue-infra ~20:29, mercury-skills-ingest ~16:09, hermes-tools-audit ~16:09) hit HTTP 503 on the inference server. This is a systemic issue — all three need retry logic in their prompts or the cron framework needs a per-job circuit-breaker that backs off when the model is unavailable.
2026-06-05 mercury-tools-audit
| Rubric | Score | Note |
|---|---|---|
| SPECIFICITY | 1/2 | Identified tool failure (terminal returning empty outputs), specific symptom, suggested root cause (broken pipe / stale PTY). |
| FOLLOW-UP | 1/2 | Clear recommendation (“Check Mercury agent’s terminal session… Re-run after restart”). |
| WIKI-INTEGRATION | 0/2 | No session file created despite detailed output. |
| SIGNAL-TO-NOISE | 1/2 | Good but the error message is verbose — feels like tool failure text leaked directly. |
| SCOPE-DISCIPLINE | 0/2 | Failed early, did not complete any audit work. |
2026-06-05 cashflow-weekly
| Rubric | Score | Note |
|---|---|---|
| ALL | 0/2 | ImportError: cannot import name 'AmbiguousJobReference' from 'cron.jobs' — Python import error, likely a code regression. This is a bug in the cron infrastructure itself, not an agent output issue. |
Action needed: The AmbiguousJobReference class was likely removed or renamed. Check recent changes to /opt/hermes/cron/jobs.py.
2026-05-14 wiki-lint-daily
| Rubric | Score | Note | | SPECIFICITY | 0/2 | Failed. Score: 0/2 | | FOLLOW-UP | 0/2 | Failed. Score: 0/2 | | WIKI-INTEGRATION | 0/2 | Failed. Score: 0/2 | | SIGNAL-TO-NOISE | 0/2 | Failed. Score: 0/2 | | SCOPE-DISCIPLINE | 0/2 | Failed. Score: 0/2 | Prompt delta suggestion: Add specific tool error handling to prompt and include retry logic in the workflow.
2026-05-14 Smart Groceries
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 daily-bill-scan
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 agent-tool-watch
| Rubric | Score | Note | | SPECIFICITY | 0/2 | Failed. Score: 0/2 | | FOLLOW-UP | 0/2 | Failed. Score: 0/2 | | WIKI-INTEGRATION | 0/2 | Failed. Score: 0/2 | | SIGNAL-TO-NOISE | 0/2 | Failed. Score: 0/2 | | SCOPE-DISCIPLINE | 0/2 | Failed. Score: 0/2 | Prompt delta suggestion: Add specific tool error handling to prompt and include retry logic in the workflow.
2026-05-14 asx-trading
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 telemetry-daily-summary
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 telemetry-12h-spotcheck
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 hermes-self-improve
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 hermes-asx-trading
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 hermes-smart-groceries
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 hermes-voice-pipeline
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 mercury-skills-ingest
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 hermes-synthetic-avatar
| Rubric | Score | Note | | SPECIFICITY | 0/2 | Failed. Score: 0/2 | | FOLLOW-UP | 0/2 | Failed. Score: 0/2 | | WIKI-INTEGRATION | 0/2 | Failed. Score: 0/2 | | SIGNAL-TO-NOISE | 0/2 | Failed. Score: 0/2 | | SCOPE-DISCIPLINE | 0/2 | Failed. Score: 0/2 | Prompt delta suggestion: Add specific tool error handling to prompt and include retry logic in the workflow.
2026-05-14 hermes-tools-audit
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 mercury-tools-audit
| Rubric | Score | Note | | SPECIFICITY | 2/2 | Score: 2/2 | | FOLLOW-UP | 2/2 | Score: 2/2 | | WIKI-INTEGRATION | 2/2 | Score: 2/2 | | SIGNAL-TO-NOISE | 2/2 | Score: 2/2 | | SCOPE-DISCIPLINE | 2/2 | Score: 2/2 | Prompt delta suggestion: Reinforce the brevity directive: ‘Output must adhere to Hermes direct tone: terse, fragments only.‘
2026-05-14 cashflow-weekly
| Rubric | Score | Note | | SPECIFICITY | 0/2 | Failed. Score: 0/2 | | FOLLOW-UP | 0/2 | Failed. Score: 0/2 | | WIKI-INTEGRATION | 0/2 | Failed. Score: 0/2 | | SIGNAL-TO-NOISE | 0/2 | Failed. Score: 0/2 | | SCOPE-DISCIPLINE | 0/2 | Failed. Score: 0/2 | Prompt delta suggestion: Add specific tool error handling to prompt and include retry logic in the workflow.