operation-logger-skill

Operation Logger

A multi-IDE skill that creates and maintains a persistent operation-log.md file, recording every step, decision, error, and file change during complex tasks.

Problem: In long conversations, earlier operations get compressed or dropped from context. The AI forgets what it already tried, repeats failed approaches, and loses track of modified files.

Solution: Treat the filesystem as long-term memory. Every meaningful operation gets recorded in operation-log.md — a single, chronological document that serves as the complete history of what happened.

Platform Support

IDE / Tool	Status	Hooks	Installation
Claude Code	Full	Native	claude plugin install operation-logger@operation-logger-skill
Cursor	Full	Pre/Post/Stop	Copy .cursor/ to project root
GitHub Copilot	Full	Session/Pre/Post/Stop	Copy .github/hooks/ to project root
Mastra Code	Full	Pre/Post/Stop	Copy .mastracode/ to project root
OpenCode	Full	Frontmatter	Copy .opencode/skills/ to ~/.config/opencode/skills/
Codex	Full	Frontmatter	Copy .codex/skills/ to ~/.codex/skills/
CodeBuddy	Full	Frontmatter	Copy .codebuddy/skills/ to ~/.codebuddy/skills/
KiloCode	Full	Frontmatter	Copy .kilocode/skills/ to ~/.kilocode/skills/
Gemini CLI	Skills	No	Copy .gemini/skills/ to ~/.gemini/skills/
Continue	Skills	No	Copy .continue/skills/ to ~/.continue/skills/
Kiro	Steering	No	Copy .kiro/ to project root
OpenClaw	Skills	No	Copy .openclaw/skills/ to ~/.openclaw/skills/
Pi Agent	Skills	No	Copy .pi/skills/ to ~/.pi/skills/
FactoryAI	Skills	No	Copy .factory/skills/ to ~/.factory/skills/
AdaL CLI	Skills	No	Copy .adal/skills/ to ~/.adal/skills/
Agent Spec	Skills	No	Copy .agent/skills/ to ~/.agent/skills/

Installation

Claude Code

claude plugin install operation-logger@operation-logger-skill

Or add to ~/.claude/settings.json:

{
  "extraKnownMarketplaces": {
    "operation-logger-skill": {
      "source": {
        "source": "github",
        "repo": "1nv1s1b1e/operation-logger-skill"
      }
    }
  }
}

Quick Install (User Commands)

For immediate /log, /log-status, /log-export access in Claude Code:

cp commands/*.md ~/.claude/commands/

Cursor

Copy the .cursor/ directory to your project root. Hooks activate automatically.

GitHub Copilot

Copy .github/hooks/ to your project root. Works with Copilot CLI, VS Code, and Coding Agent.

Kiro

Copy .kiro/ to your project root. Steering files provide persistent guidance.

Other Platforms

Copy the corresponding .{platform}/skills/operation-logger/ directory to your platform's skills location.

Usage

Slash Commands (Claude Code)

Command	Description
/log	Start operation logging for the current task
/log-status	Show current log status and recent operations
/log-export	Export statistics summary from the log

Auto-Triggering

The skill triggers for multi-step tasks involving:

• Multi-file refactoring or codebase migration
• Complex debugging sessions
• Building features spanning multiple modules
• Security audits, performance optimization
• Any work touching 3+ files

Example Log Entry

#### [OP-007] EDIT | Add validation to PUT endpoint
- **Action**: Added express-validator middleware to PUT /api/users/:id/profile
- **Result**: Validation now rejects invalid input with 400 status
- **Files**: src/routes/profile.ts (modified)
- **Context**: Resolves OP-005 (missing input validation)

Log Structure

The operation-log.md has three sections:

1. Running Summary — Live-updated snapshot of current state (updated every 5 ops)
2. Session Entries — Each conversation gets its own section
3. Operation Entries — Chronological record with type, action, result, files, context

Operation Types

SEARCH READ EDIT CREATE DELETE TEST BUILD COMMAND DECISION ERROR FIX RESEARCH PLAN REVIEW

Hook Behavior

For platforms with hook support, the operation-logger provides 4 lifecycle hooks:

Hook	Trigger	Action
userPromptSubmit / sessionStart	Every user message / session start	Remind to read operation-log.md if it exists
preToolUse	Before Write/Edit/Bash/Read	Inject Running Summary into context
postToolUse	After Write/Edit	Remind to log the operation
stop / agentStop	Agent tries to stop	Check Final Summary is filled; continue if not

Scripts