Stores durable facts in a categorized, plain-markdown vault on disk, alongside your agent's built-in memory.
Coding
reconcile
Try itTripwire check for multi-session drift. Scans state files, recent commits, and file conflicts caused by parallel Claude Code sessions.
What it does
<!-- x-source: agent-workspace/SKILL.md @ 7b7211e -->
The skill document
/reconcile — Multi-Session Drift Check
When running multiple Claude Code sessions in parallel (especially with worktrees), files can drift out of sync. This skill scans for inconsistencies and proposes fixes.
Invocation: deliberately model-invocable — the scan is read-only and auto-triggering when drift smells is the point. Anything that changes files is proposed, never applied.
When to Use
- After merging worktree branches back to the default branch
- When something "feels off" after parallel work
- After a crash where multiple sessions were open
- As a periodic sanity check during heavy parallel workflows
Instructions
0. Detect the default branch
git rev-parse --git-dir >/dev/null 2>&1 || { echo "not a git repository — nothing to reconcile"; exit 0; }
DEFAULT=$(git symbolic-ref --short refs/remotes/origin/HEAD 2>/dev/null | sed 's|^origin/||')
DEFAULT=${DEFAULT:-main} # no remote configured — fall back and say so in the report
1. Scan recent commit history across all branches
git log --all --oneline --since="24 hours ago" --graph
Look for:
- Multiple branches touching the same files
- Commits on branches that haven't been merged
- Conflicting changes (same file modified differently on different branches)
2. Check for file-level conflicts
For each file modified on multiple branches, diff the versions:
git diff "$DEFAULT".. --
Flag files where:
- Both branches modified the same lines (merge conflict risk)
- One branch deleted what another modified
- Timestamps or "Last Updated" fields diverged
3. Check state file consistency
If your project uses state files (TODO lists, priority files, decision logs, session logs), check:
- Duplicate entries: Same task or decision logged twice from different sessions
- Contradictory state: One session marked something complete, another added it as in-progress
- Timestamp drift: "Last Updated" dates that don't match the most recent actual edit
- Orphaned references: Files or sections referencing things that were removed in another session
4. Check for SSOT violations
If your project has single-source-of-truth rules (facts that should only live in one file):
- Scan for duplicated facts across files
- Flag any case where the same information appears in multiple locations with different values
- Check that cross-references point to files that still exist
5. Report
RECONCILE — [DATE]
BRANCHES CHECKED:
- [branch list with last commit date]
FILE CONFLICTS:
- [file] — modified on [branch1] and [branch2] — [conflict type]
STATE DRIFT:
- [issue description]
SSOT VIOLATIONS:
- [duplicated fact] — found in [file1] and [file2]
PROPOSED FIXES:
1. [fix description]
OVERALL: [CLEAN / N issues found]
6. Apply fixes (with approval only)
Present each fix individually. Wait for approval before applying. Common fixes:
- Merge the newer version of a conflicting file
- Remove duplicate entries (keep the more detailed one)
- Update timestamps to match actual last-edit dates
- Resolve SSOT violations by keeping the canonical source and updating references
Design Principles
- Read-only by default. Report, wait for approval.
- Prefer evidence of intent over timestamps. When two versions conflict, look at which change the surrounding work depends on (commit messages, linked edits, whether other files reference the new value). A stale session can easily produce the newer timestamp. Use recency only as a tie-breaker when intent is unreadable.
- Preserve intent. Don't auto-resolve — different sessions may have had different goals.
- Fast. Git commands and file reads only. Under 30 seconds for a full scan.
Related skills
Find why your productivity system keeps failing, then apply the smallest fix — capacity math, bottleneck routing, durable local notes.
Run Git operations — commits, branches, merges, rebases, conflict resolution, and recovery — with safety rules enforced.
Generate and edit Draw.io, Mermaid, and Excalidraw diagrams from natural language using a structured JSON spec.
Join a video meeting as an AI bot with voice, avatar, and screenshare across four operating modes.
Write, debug, and tune Playwright specs with locator strategy, trace diagnosis, and CI-aware timeouts.
More from conorbronsdon
Browse all skillsAudit and rewrite text to remove AI-generated writing patterns, with detect-only and edit-in-place modes.
Pre-investment research and deal-memo generation for angel investors. Use when asked to "diligence [company]", "research this startup for an investment", "build a deal memo", or to prep for a founder call. Runs parallel web research with strict citation rules and produces a nine-section memo that separates verified facts from claims. Never outputs an invest/pass recommendation.
Multi-agent deep review for code PRs in any repo. Use when asked to "deep review this PR," "multi-agent review," "review
Audit an LLM evaluation or benchmark repo for integrity and credibility practices. Use when asked to "audit my benchmark," "is my eval trustworthy," "check my leaderboard for contamination," "review this benchmark's methodology," or "what would a reviewer attack in my eval." Greps the target repo for evidence across seven dimensions (pre-registration, contamination, holdout hygiene, judge validity, statistical honesty, reproducibility, leaderboard exclusions) and emits a scored report with file:line evidence, severity, and concrete fixes.
Researches and maps a guest’s podcast appearance history to identify their last topics, unclaimed angles, and receptiveness for targeted outreach.
Structured session lifecycle for Claude Code — start, checkpoint, end, and daily heartbeat commands that maintain project state across conversations.