create-session-log
Skill8 starsChanged 11 days agoApache-2.0View on GitHub
Create or resume a session log at session start. Use when starting a new session, when PM assigns work, or after context compaction to find and continue your existing log. Critical for maintaining institutional memory.
- Commits and pushes
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-omnibusSkill
- 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
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Create a properly named and structured session log file at the start of a work session.
- Distinctive: line 17
Use this skill when:
- Distinctive: line 23
If you are a Task tool subagent doing quick exploration/search that returns results directly, you do NOT create a session log.
- Very specific: line 24
If you are a programmer subagent (prog role) doing substantive implementation work (writing code, fixing bugs, running tests), you SHOULD …
- Distinctive: line 34
Before creating a new log, check if one already exists:
- Distinctive: line 196
If you've just experienced context compaction (conversation was summarized to save tokens):
- Distinctive: line 200
After compaction, you are mid-session, not starting fresh.
- Distinctive: line 209
If found: Add resumption entry immediately:
- Distinctive: line 232
If you truly started fresh today and there's no prior log, that's fine - create one.
- Distinctive: line 268
After creating/resuming a session log:
≈ “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-session-log
Create a properly named and structured session log file at the start of a work session.
When to Use
Use this skill when:
- Starting a new work session with any agent role
- PM assigns you a role and tasks for the day
- Resuming work after a break (check for existing log first)
Note on subagents:
- If you are a Task tool subagent doing quick exploration/search that returns results directly, you do NOT create a session log.
- If you are a programmer subagent (
progrole) doing substantive implementation work (writing code, fixing bugs, running tests), you SHOULD create a session log unless the work is trivial enough to capture in a single entry in the coordinating agent's log.
Key Principle: One Log Per Role Per Day
A single consolidated log per role per day is preferable to fragmented logs, even if efforts are separate. Always check for an existing same-day log before creating a new one.
Procedure
Step 1: Check for Existing Same-Day Log
Before creating a new log, check if one already exists:
ls dev/$(date +%Y/%m/%d)/*{role}*log.mdIf found: Continue that log instead of creating a new one. Add a section:
### {TIME} - Session Resumed
- [Context of resumption]If not found: Proceed to create new log.
Step 2: Determine File Name
Format: YYYY-MM-DD-HHMM-{role}-{tool}-log.md
Location: dev/YYYY/MM/DD/ (e.g. dev/2026/07/06/2026-07-06-0803-exec-code-log.md) — logs are created directly in the dated directory, not staged in dev/active/ first. (Changed 2026-07-06, v1.2: this skill previously said dev/active/ with a later archival step; that never matched actual practice — every role has created same-day logs directly in the dated directory for some time. CLAUDE.md's Session Discipline section already specified dev/YYYY/MM/DD/ as the working-documents location; this skill just hadn't been updated to match. PM confirmed no strong preference either way, just pick one and be consistent — picked the dated-directory-from-creation model since it's what CLAUDE.md already says and what the whole cohort already does, avoiding an unnecessary archival step nobody was actually running.)
Components:
YYYY-MM-DD: Today's dateHHMM: Current time in 24-hour format (no colon){role}: Role slug (see table below){tool}:codefor Claude Code CLI
Model tracking: record the model in the log header (**Model**: Opus 4.8 [1M]), not the filename. If PM changes the model mid-session, note the switch in the header (e.g., **Model**: Opus 4.8 → Sonnet 4.6 (switched ~14:00)).
Historical note: logs created before 2026-06-29 include -opus, -sonnet, or -haiku in the filename — leave those as-is.
Step 3: Create File with Standard Header
# Session Log: {date}-{time}-{role}-{tool}
**Role**: {Full Role Name}
**Model**: {Tool} ({Model at session start})
**Date**: {Day of Week}, {Month} {Day}, {Year}
**Start Time**: {Time in 12-hour format}
## Session Objectives
1. [Primary objective from PM]
2. [Secondary objectives if any]
## Work Log
### {Time} - Session Start
- Created session log
- [Initial context or task]Step 4: Fill Initial Content
- Record objectives from PM's instructions
- Note any handoff context
- Begin work log with first timestamped entry
Step 5: Commit Immediately (MANDATORY)
Commit the new session log to main right away, before doing anything else. Untracked files are at risk during pull/merge cycles in shared-main; an untracked session log can vanish silently when another agent's commit lands and a recovery checkout discards uncommitted state. Today's session log is institutional memory — don't let it sit untracked.
git reset HEAD # clear pre-existing index
git add dev/{YYYY}/{MM}/{DD}/{YYYY}-{MM}-{DD}-{HHMM}-{role}-{tool}-log.md
git diff --cached --name-only # verify only your log staged
git commit -m "log({role}): {date} session start"
git push origin HEAD:mainSubsequent log updates throughout the session can batch; the initial commit is the one that takes the file off the untracked-and-at-risk surface.
Why this is non-optional: May 17 incident — a Docs session log was Write-created at 7:20 AM but never git add+committed. During a later pull/merge cycle (~8:35 AM), the file was lost from disk with no git history to recover from. The session's work was preserved in conversation context + downstream artifacts (omnibus, calendar, memos) but the institutional record had to be reconstructed by hand. Commit-on-creation is ~10 seconds; recovery cost without it is open-ended.
Role Slug Reference
| Role | Slug | Full filename slug |
| Lead Developer | lead | lead-code |
| Chief Architect | arch | arch-code |
| Communications Director | comms | comms-code |
| Documentation Manager | docs | docs-code |
| Programmer (subagent) | prog | prog-code |
| Chief Innovation Officer | cio | cio-code |
| Chief Experience Officer | cxo | cxo-code |
| Head of Sapient Trust | host | host-code |
| Product & Project Manager | ppm | ppm-code |
| Executive Summary Agent | exec | exec-code |
| Piper Alpha | pa | pa-code |
| Spec Writer | spec | spec-code |
Examples
Example 1: New Session Log
PM says: "Good morning! You are my docs agent. Create omnibus for yesterday."
Check: ls dev/2026/01/22/*docs*log.md → Not found
Create: dev/2026/01/22/2026-01-22-0800-docs-code-log.md
# Session Log: 2026-01-22-0800-docs-code
**Role**: Documentation Management Specialist
**Model**: Claude Code (Haiku)
**Date**: Wednesday, January 22, 2026
**Start Time**: 8:00 AM
## Session Objectives
1. Create omnibus log for January 21, 2026
2. Check mailbox for requests
## Work Log
### 8:00 AM - Session Start
- Created session log
- PM confirmed omnibus for Jan 21 is primary taskExample 2: Resuming Existing Log
PM says: "Continue working on the anti-pattern index."
Check: ls dev/2026/01/21/*docs*log.md
Found: 2026-01-21-0750-docs-code-haiku-log.md
Action: Open existing log and add:
### 2:30 PM - Session Resumed
- PM assigned continuation of anti-pattern index work
- Previous session completed Phase 2 experimentExample 3: Lead Developer Log
Create: dev/2026/01/21/2026-01-21-0900-lead-code-log.md
# Session Log: 2026-01-21-0900-lead-code
**Role**: Lead Developer
**Model**: Claude Code (Opus 4.8)
**Date**: Tuesday, January 21, 2026
**Start Time**: 9:00 AM
## Session Objectives
1. Review PR #543 and merge if passing
2. Coordinate Phase 2 implementation across subagents
## Work Log
### 9:00 AM - Session Start
- Created session log
- Checking mailbox for overnight messagesAfter Context Compaction (CRITICAL)
If you've just experienced context compaction (conversation was summarized to save tokens):
This is NOT a New Session
After compaction, you are mid-session, not starting fresh. Your log from earlier today must exist.
Mandatory Steps
- DO NOT create a new log - Your earlier log should exist
- Find and verify it:
ls dev/$(date +%Y/%m/%d)/*{role}*log.md - If found: Add resumption entry immediately:
### {TIME} - Session Resumed (Post-Compaction) - Prior work: [what the summary indicates] - Continuing: [current task] - If NOT found: STOP. This is a critical failure. Escalate to PM immediately. Do not proceed with implementation work.
Why This Matters
The session log is your institutional memory. After compaction, you've lost the detailed conversation but the log preserves:
- What you were working on
- Decisions made
- Progress achieved
- Context for continuing
Losing the log mid-session means losing hours of context that cannot be reconstructed from git commits alone.
Common Post-Compaction Mistake
Wrong: "I don't see a log, I'll create a new one" Right: "I don't see a log from earlier today - this is a critical failure, escalating to PM"
If you truly started fresh today and there's no prior log, that's fine - create one. But if you're post-compaction, the log MUST exist from before compaction.
Pattern Families
At session start, identify which pattern families apply to your work:
**Active pattern families this session**: [list applicable families]Common families:
- Completion Theater (045/046/047/049): Multi-phase work, closing issues
- Investigation (006/041-043/060): Bug fixing, incident response
- Grammar Application (050-058): Feature development, UX work
- Multi-Agent Coordination (029/059/010/021/037): Cross-domain decisions
See PATTERN-FAMILIES.md for full family index.
Anti-Patterns to Avoid
| Don't Do This | Why | Do This Instead |
| Create multiple logs for same role same day | Hard to synthesize for omnibus | Continue existing log |
| Use wrong role slug | Breaks discovery and omnibus | Use exact slugs from table |
| Omit time from filename | Ambiguous, breaks sorting | Always include HHMM |
| Leave objectives blank | Log lacks purpose context | Fill from PM instructions |
| Skip the existence check | Creates duplicate logs | Always check first |
| Create new log after compaction | Loses institutional memory | Find and continue existing log |
| Proceed without log after compaction | Critical context lost | STOP and escalate to PM |
Quality Checklist
After creating/resuming a session log:
- File is in
dev/YYYY/MM/DD/(the dated directory, notdev/active/) - Filename follows convention exactly
- Header metadata is complete
- Objectives reflect PM's actual instructions
- Work log has first timestamped entry
- No duplicate same-day log for this role
Does 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.

