Pi Workflow Orchestration
PassAudited by ClawScan on May 1, 2026.
Overview
This is a coherent workflow/self-improvement skill, but it intentionally creates persistent lesson context and can add a session-start reminder hook, so users should review what gets saved or promoted.
This skill appears safe to install if you want a persistent workflow and lesson-capture system. Before enabling hooks or syncing lessons, verify hook status, use dry-run previews, redact secrets from lesson/error files, and require explicit review before promoting lessons into AGENTS.md, SOUL.md, or TOOLS.md.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Future sessions may be influenced by saved lessons or session notes, and those notes may contain project details or mistaken guidance.
The skill explicitly persists lessons and reuses them in later sessions, which is coherent with self-improvement but can preserve sensitive, incorrect, or over-broad rules if not reviewed.
After ANY correction from the user: update `tasks/lessons.md` ... Review lessons at session start for relevant projects
Keep lesson entries concise, avoid secrets, periodically prune outdated content, and require user review before promoting lessons into long-term agent instruction files.
New sessions may automatically receive workflow reminders that steer the agent toward lesson logging and self-improvement behavior.
The included hook is designed to run at future session starts and inject a virtual reminder. It does not write files, but it is persistent behavior that affects agent context.
The hook fires automatically on every agent bootstrap (session start).
Check `openclaw hooks list` after installation and disable `pi-workflow` if you do not want bootstrap reminders.
The agent may be more proactive about analysis, tests, and code changes when a task appears to fit the workflow.
The skill encourages delegated analysis and autonomous bug fixing. This fits the workflow purpose, but users should notice that it may reduce back-and-forth before the agent takes action.
Use subagents liberally ... When given a bug report: just fix it. Don't ask for hand-holding
For sensitive repositories or production-impacting changes, ask the agent to present a plan and wait for approval before editing, running costly commands, or making irreversible changes.
Sensitive project notes or bad workflow rules could be pushed to a repository and reused by other installations.
The documented sync workflow can propagate local lessons into a version-controlled skill reference, which is useful but can spread private or flawed lessons if not curated.
Lessons are version-controlled on GitHub ... Can be shared with other instances of the skill
Use the dry-run mode, review and redact lessons before committing, and avoid syncing workspace-specific secrets or private implementation details.