create-omnibus
Skill8 starsChanged 11 days agoApache-2.0View on GitHub
Create an omnibus session log synthesizing all agent sessions from a given date. Use at the start of each Docs session (omnibus before new work), or when PM says "create omnibus", "omnibus for yesterday".
In this repository39 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .cursor/rules/architecture-patterns.mdcCursor rule
- .cursor/rules/completion-discipline.mdCursor rule
- .cursor/rules/github-tracking.mdcCursor rule
- .cursor/rules/programmer-briefing.mdcCursor rule
- .cursor/rules/verification-first.mdcCursor rule
- ab-a-isolationSkill
- assign-sprint-safelySkill
- audit-cascadeSkill
- brief-coding-agentSkill
- check-mailboxSkill
- cleanup-dev-activeSkill
- close-issue-properlySkill
- cohort-attention-rollupSkill
- compost-reviewSkill
- continue-narrativeSkill
- create-session-logSkill
- cut-releaseSkill
- delete-module-safelySkill
- deliver-mailSkill
- discovered-work-captureSkill
- doc-sync-sweepSkill
- draft-blog-postSkill
- draft-weekly-shipSkill
- duty-cycle-tickSkill
- narrative-verificationSkill
- piper-draft-issueSkill
- piper-draft-specSkill
- piper-sprint-planSkill
- piper-stakeholder-updateSkill
- piper-synthesize-feedbackSkill
- propose-featureSkill
- publish-to-blogSkill
- query-github-boardSkill
- template-auditSkill
- trust-checkSkill
- update-calendarSkill
- update-current-stateSkill
- update-piperSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 11 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 11
Create an omnibus session log synthesizing all agent sessions from a given date.
- Distinctive: line 16
When PM says "create omnibus", "omnibus for yesterday", etc.
- Distinctive: line 17
After any HIGH-COMPLEXITY day (4+ agents)
- Distinctive: line 29
What this means for the synthesizer (Docs): keep the daily synthesis quality high.
- Distinctive: line 35
MANDATORY FIRST STEP: Read the core methodology before doing anything else:
- Distinctive: line 50
Ask PM if not obvious. Default: previous calendar day.
- Distinctive: line 118
PASS: every mentioned role has a log, or PM has confirmed which mentioned roles are genuinely not-active that day.
- Distinctive: line 133
When a session log makes a factual claim about another agent's work (e.g., "Lead Dev shipped X," "CXO endorsed Y," "Architect filed Z"), …
- Distinctive: line 134
If the claim and the referenced log diverge, flag the discrepancy in the omnibus rather than picking a side.
- Distinctive: line 229
Run through the validation checklist from the methodology:
- Very specific: line 286
If the second command lists a session log, it was staged in the wrong place: mv it to its own dev/YYYY/MM/DD/ (derived from the filename …
- Very specific: line 306
Use Python csv.writer with quoting=csv.QUOTE_MINIMAL, lineterminator='\n' to handle commas in notes and enforce LF endings (.gitattributes …
- Distinctive: line 308
Commit + push as a separate commit immediately after the omnibus commit.
- Very specific: line 316
Verification: after append, wc -l docs/internal/operations/agent-activity-log.csv should increase by exactly the number of new rows.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
create-omnibus
Create an omnibus session log synthesizing all agent sessions from a given date.
When to Use
- At the start of each Docs session (omnibus before new work)
- When PM says "create omnibus", "omnibus for yesterday", etc.
- After any HIGH-COMPLEXITY day (4+ agents)
Role of the Omnibus (Code-era reframing, 2026-04-27)
The omnibus's role shifted with the Apr 22–26 leadership migration to Code. It is no longer the primary input for weekly workstream reviews — those reviews now read primary session logs directly (filesystem access in Code makes a 7-day session-log read nearly as fast as one omnibus, with materially higher fidelity).
The omnibus's standing functions going forward:
- Daily narrative arc — readable cross-role view of a day's work; useful for blog posts, cross-day pattern detection, retrospective reading.
- Coverage check — workstream-review authors scan the omnibus after writing their primary-sourced memo, to flag any role-relevant content the omnibus missed (becomes the omnibus's standing quality feedback loop).
- Continuity for slower-cadence consumers — anyone catching up after a multi-day gap prefers the condensed view to a 30-log read.
What this means for the synthesizer (Docs): keep the daily synthesis quality high. The omnibus is no longer "the input" but it remains a referenced surface — coverage-check feedback from workstream-review authors will surface omissions; treat those as amendment candidates.
What this does NOT mean: the omnibus is not "deprecated" or "lower-priority." It is repurposed. Same daily cadence; same structural rigor; different consumers' read-pattern.
Pre-Flight: Load the Methodology
MANDATORY FIRST STEP: Read the core methodology before doing anything else:
# Read this EVERY TIME — do not work from memory
cat docs/internal/development/methodology-core/methodology-20-OMNIBUS-SESSION-LOGS.mdThe methodology defines format selection, the 6-phase method, line limits, compression ratios, timeline rules, executive summary format, validation checklist, and common pitfalls. Follow it precisely.
This skill is a runbook for executing the methodology, not a replacement for it.
Procedure
Step 1: Determine Target Date
Ask PM if not obvious. Default: previous calendar day.
# Set target date
TARGET="YYYY-MM-DD"
YEAR="YYYY" MONTH="MM" DAY="DD"Step 2: Source Discovery (Methodology Phase 1)
# Find all session logs for target date
ls dev/$YEAR/$MONTH/$DAY/*log*.md
# Count them
ls dev/$YEAR/$MONTH/$DAY/*log*.md | wc -l
# Check for cloud agent artifacts in mailboxes
find mailboxes/*/read/ -name "*$TARGET*" -type f
# Check dev/active for same-day items
find dev/active/ -name "*$TARGET*" -type fReport inventory to PM:
Found N session logs for [date]:
- [list each with role and filename]
- [note any cloud-only agents with memos but no session logs]Step 2.5: Cross-Reference Gate (MANDATORY — added 2026-04-22 after drift incident)
Before proceeding to format selection, verify the source-log set is complete by cross-referencing agent mentions inside each log against the roles represented in the source set.
The failure this prevents: synthesizing an omnibus from "the logs currently in tree" without checking whether those logs mention other agents whose own logs haven't been downloaded yet. This is Pattern-062 (Assembly Assumption) applied to omnibus synthesis — individually-correct source logs can produce a collectively-incomplete omnibus if an agent's reference to "I got a memo from PPM" is present in three other logs but PPM's own log is absent from the source set.
Procedure:
- Enumerate the roles present in the source set — build a set
{Lead Dev, Docs, CXO, ...}based on each source log's role. - Scan each source log for mentions of other agent roles — grep each log for role names and agent patterns:
# Known agent role vocabulary (keep in sync with the actor names list in Step 5)
AGENTS="Lead Dev|Lead Developer|Docs|Documentation Management|docs-code|PA|Piper Alpha|CXO|Chief Experience|CIO|Chief Innovation|PPM|Principal Product|Architect|Chief Architect|arch|Comms|Communications|HOST|Exec|Chief of Staff|code-opus|Code Agent"
for log in dev/$YEAR/$MONTH/$DAY/*log*.md; do
echo "=== $(basename $log) ==="
grep -oE "($AGENTS)" "$log" | sort -u
done- Compile the union of all mentioned roles across the source set.
- Compare mentioned-set against source-set:
- Any role mentioned but not in source-set → flag as potential missing log
- For each flagged role, ask PM: "I see mentions of [ROLE] in today's logs — was [ROLE] actually active today, or are these backreferences to prior-day work?"
- Also check for cross-role artifacts: scan
dev/active/anddev/$YEAR/$MONTH/$DAY/for artifacts (non-log files) dated on the target date. If an artifact is attributed to a role whose session log is not in the source set, that role was working — the session log may be missing.
# Check dev/active/ for artifacts dated on target
find dev/active/ dev/$YEAR/$MONTH/$DAY/ -name "*$TARGET*" -type f | grep -v "log.md$"
# For each, check the file's authorship line (often "Author: ROLE" or "Prepared by: ROLE")- Gate decision:
- PASS: every mentioned role has a log, or PM has confirmed which mentioned roles are genuinely not-active that day. Proceed to Step 3.
- FAIL: a mentioned role's log is likely missing and downloadable. STOP — ask PM to download (if Chat) or file (if Code) the missing log before proceeding. Synthesizing without it would produce a drifted omnibus, as happened on 2026-04-19 with the Apr 16 omnibus (later amended 2026-04-22 after PPM, CIO, and HOST 4/16 logs were downloaded).
If the gate fails and PM declines to fetch (e.g., the agent truly is reachable only later): document the gap explicitly in the omnibus Sources section — "NOTE: [ROLE] session log not available at synthesis time; content inferred from cross-references only. May require amendment when log becomes available."
Do not paper over the gap by proceeding as if the source set were complete.
Step 2.6: Cross-Role Mentions Verification (added 2026-04-27 per CXO ask)
After source set passes Step 2.5, verify that cross-role assertions in one log are consistent with the other log they reference.
The failure this prevents: PA's session log says "CXO scored S2 at 9/9" — but CXO's session log doesn't mention scoring S2 at all. Both logs are "present" (Step 2.5 passes), but they disagree on facts. Pattern-062 at the assertion layer.
Procedure (do for high-impact cross-role assertions, not exhaustively):
- When a session log makes a factual claim about another agent's work (e.g., "Lead Dev shipped X," "CXO endorsed Y," "Architect filed Z"), spot-check the referenced agent's session log for the same event.
- If the claim and the referenced log diverge, flag the discrepancy in the omnibus rather than picking a side. Sample framing: "PA's log records X; CXO's log records Y. Discrepancy preserved for resolution."
- The discipline is "name the divergence," not "force consensus." Both reads may be true at different times (yesterday's mail-discipline cascade is the canonical case — racing snapshots, not disagreement).
Skip this step for: routine references (cross-role mail receipts, attendance), well-aligned multi-role threads (Phase E scoring chains where everyone explicitly references the same artifacts).
Step 3: Format Selection
Based on session count and characteristics:
| Criteria | Format | Line Limit |
| 1 session, single goal | MINIMAL | ~50 lines |
| 1-2 agents, single goal | STANDARD | 300 lines |
| 3+ parallel streams, architectural decisions, coordination | HIGH-COMPLEXITY | 600 lines |
Justify your selection in the opening paragraph. If borderline, ask PM.
Step 4: Read ALL Source Logs (Methodology Phase 2)
Read every log completely. No skimming.
For each log, extract:
- Every timestamped entry
- Cross-references to other agents (handoffs, mentions)
- Reflective content (especially Lead Dev end-of-session)
Step 5: Build Timeline (Methodology Phases 3-5)
The timeline is non-negotiable. It must show events from ALL agents interleaved by time.
Format by day type:
- STANDARD: Simple bullet list with bold actor names
- HIGH-COMPLEXITY: Phase-grouped with time period headers
Rules (enforced strictly):
- Each entry: 1-2 lines MAX
- Interleaved by time, not grouped by agent
- Preserve coordination handoffs as distinct entries
- Preserve causality chains (discovery → decision → implementation)
- Bold actor names consistently
Actor names (use these, not slugs):
- xian — PM/founder
- Lead Developer — code-side dev coordination
- Documentation Management or docs-code — omnibus, mailbox, blog pipeline
- Chief Architect — ADRs, architecture
- Chief of Staff — cross-workstream synthesis
- HOST — agent welfare, human network
- Communications Chief — blog, narrative
- CXO — UX testing, Colleague Test
- CIO — methodology, patterns
- PPM — sprint planning, roadmap
Step 6: Write Executive Summary (Methodology Phase 6)
4 sections, each with terse 1-line bullets:
- Core Themes (3-5 bullets) — major accomplishments, breakthroughs, coordination patterns
- Technical Details (5-8 bullets) — specific implementations, architecture decisions, infrastructure
- Impact Measurement (4-6 bullets) — quantitative metrics, qualitative improvements
- Session Learnings (5-8 bullets) — what worked, what caused friction, patterns to replicate/avoid
No paragraphs. Each bullet = one concise line. Source logs have details.
Step 7: Verify Canonical References (MANDATORY when applicable)
If the day includes a ratified PDR, ADR, Pattern, or methodology doc, the omnibus must quote the canonical source — not paraphrase from memory or from session log summaries.
The rule: quote or reference — never paraphrase canonical content.
Companion principle (added 2026-04-27 per CXO ask): verify at point of creation, not downstream. When first citing a canonical artifact in an omnibus draft, open the doc and confirm titles/principle-names verbatim — don't trust the previous omnibus's wording or your own memory. Each omnibus is a fresh authoring surface; paraphrase drift accumulates if each generation lifts from the prior generation.
For every mention of a canonical artifact (PDR-XXX, ADR-XXX, Pattern-XXX, methodology-XX, PDR/ADR/Pattern by name):
- Open the canonical doc at its authoritative path:
- PDRs:
docs/internal/product/pdr/PDR-XXX-*.md - ADRs:
docs/internal/architecture/adrs/ADR-XXX-*.md - Patterns:
docs/internal/architecture/patterns/pattern-XXX-*.md - Methodologies:
docs/internal/development/methodology-core/methodology-XX-*.md
- PDRs:
- Copy principle names, titles, and key terms verbatim from the canonical doc into the omnibus entry.
- If the doc is unavailable (not yet committed, or path uncertain): do not invent or paraphrase — write "PDR-XXX ratified; titles to be confirmed" and flag for correction.
This prevents the class of error where successive agents paraphrase the same canonical content and it drifts over time.
Examples of what to do:
✅ "PDR-004 delivered. Four principles: (1) The Session Belongs to the User, (2) Offer-First Activation, (3) Piper Coordinates Understanding, (4) The LLM Floor Guarantee." (principle names copied verbatim from PDR-004)
✅ "PDR-004 delivered — four principles governing ongoing experience; see doc for details." (referenced without summarizing content)
❌ "PDR-004 delivered. Four principles: presence over performance, specificity as care, honest boundaries, growth through use." (paraphrased from memory — the canonical principle names are different)
Step 8: Validate (Methodology Checklist)
Run through the validation checklist from the methodology:
Timeline (non-negotiable):
- Unified chronological timeline EXISTS
- Events from ALL agents interleaved by time
- Coordination handoffs visible as distinct entries
- Causality chains preserved
Canonical references:
- Every PDR/ADR/Pattern/methodology mention uses verbatim canonical names (Step 7)
- No paraphrased principle names, titles, or key terms
Format & Quality:
- All sessions identified and read completely
- Format selection justified
- LINE COUNT UNDER LIMIT (300 Standard / 600 High-Complexity)
- Timeline entries 1-2 lines max
- Executive summary bullets 1 line max
- Actor names consistent and bold
Additional for HIGH-COMPLEXITY:
- All parallel work streams captured distinctly
- Phase groupings reflect actual work patterns
- Handoff moments preserved
- Strategic pivots captured
Step 9: Write the File
# Create omnibus file
# File: docs/omnibus-logs/YYYY-MM-DD-omnibus-log.mdHeader format:
# Omnibus Log: [Month Day, Year]
**Day**: [Day of week]
**Sessions**: [N] ([list roles])
**Day Type**: [STANDARD/HIGH-COMPLEXITY] — [brief descriptor]
**Justification**: [Why this format. What made the day complex or standard.]
**Git Commits**: [count or "N+"]Step 10: Confirm Source Logs Are in Their Dated Home (Final Step Before Reporting)
There is no archive move. Session logs are created directly in dev/YYYY/MM/DD/ (create-session-log v1.2, 2026-07-06), so the sources are already where they belong once the omnibus is written. (The old step moved logs from dev/active/; nothing stages logs there, so it was dead. Removed 2026-10-08, defect C9 from CIO's R6 step 4 routing.)
# Count the target date's logs in their dated home (the number for the Step 11 report)
ls dev/$YEAR/$MONTH/$DAY/*-log.md 2>/dev/null | wc -l
# Anything stranded in dev/active/ that IS a session log (should be none)
ls dev/active/20*-log.md 2>/dev/nullIf the second command lists a session log, it was staged in the wrong place: mv it to its own dev/YYYY/MM/DD/ (derived from the filename date), and say so in the report.
Step 10.5: Activity-Log Reconciliation (Shape B per Janus 3-layer architecture)
MANDATORY when omnibus is committed: append PM-side rows for the omnibus's covered date(s) to docs/internal/operations/agent-activity-log.csv. Per the Janus-preferred Shape B integration (PM-endorsed 2026-05-09, Docs pick-disposition 2026-05-12): one row per role session log captured in the omnibus, plus reconstructed rows for any mail-only / web-side role activity captured in the omnibus's cross-reference gate.
CSV schema: date,role,slug,environment,model,log_filename,notes
Procedure:
- For each session-log file enumerated in the omnibus's Sources block, append one row:
date= the omnibus target date (YYYY-MM-DD)role= canonical role name (Lead Dev / Docs mgr / Piper Alpha / Chief Architect / HOST / Chief Innovation Officer / Chief Experience Officer / Communications / Principal PM / Programmer subagent / Code agent (special assignment))slug= the role slug from the filename (lead,docs,pa,arch,host,cio,cxo,comms,ppm,exec,prog,code)environment=codefor Code sessions;webfor chat-side (rare post-migration)model=opus(default; check filename if unsure)log_filename= basename of the session lognotes= a one-line factual summary of the day's substantive work for that role (issue numbers + memo subjects + ship deliverables; pulled from the omnibus's Core Themes section)
- For mail-only / web-side activity captured via the cross-reference gate (e.g., "Exec active by mail, no local log; reconstructed from outbound to 11 inboxes"), append one row with role = canonical name, slug = best-fit, environment =
web,log_filename= empty or descriptor likeweb-mail-only, notes = brief description. - Use Python
csv.writerwithquoting=csv.QUOTE_MINIMAL, lineterminator='\n'to handle commas in notes and enforce LF endings (.gitattributesenforces LF on checkout, butcsv.writerdefaults to CRLF — specify explicitly). Neverecho >>— corrupts CSV. - Commit + push as a separate commit immediately after the omnibus commit; suggested message:
docs(activity-log): {date} PM-side rows appended (Shape B reconciliation) N rows for {date} sessions: {role-list}. Per Janus 3-layer architecture; cross-project aggregator pulls from this canonical.
Verification: after append, wc -l docs/internal/operations/agent-activity-log.csv should increase by exactly the number of new rows. tail -N to spot-check.
Why this is Shape B not Shape A: the row-append is a separate step (not folded into the omnibus write) so retrospective backfills (e.g., a multi-day catch-up) use the same code path as single-day updates. Janus's stated preference; Docs's pick-disposition (memo to Janus 2026-05-12). The architectural separation also means activity-log updates can run ad-hoc (without an omnibus) when useful.
Step 11: Report to PM
Omnibus complete for [date]:
- Format: [STANDARD/HIGH-COMPLEXITY]
- Sessions covered: N
- Line count: N (limit: N)
- Source logs in dev/YYYY/MM/DD/: N files (stranded in dev/active/: N)
- Activity-log rows appended: N (Shape B reconciliation)
- Key themes: [2-3 sentence summary]Common Pitfalls (from methodology)
- The "Main Log" Trap — never assume one log tells the whole story
- The Memory Shortcut — always extract from actual timestamped entries, not memory
- The Sessions Table Substitution — a sessions overview table is NOT a timeline
- The Detail Bloat — source logs have details; omnibus is synthesis
- Skipping the methodology — re-read
methodology-20every time, not just the first time
Quick Reference
docs/internal/development/methodology-core/methodology-20-OMNIBUS-SESSION-LOGS.md
↑ THE SOURCE OF TRUTH — read before every omnibusDoes it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

