# AI use cases for Penelopa AI (chigwell/Penelopa.ai) Repo: [chigwell/Penelopa.ai](https://github.com/chigwell/Penelopa.ai) | Use cases: 4 | Page: https://dedupedai.com/use-cases/chigwell-penelopa-ai/ Penelopa.ai is a JavaScript tool that captures real Codex and Claude Code sessions, finds repeated workflow patterns, and turns them into reusable skills, checks, prompts, and recommendations. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Connect Codex and Claude Code sessions to Penelopa: https://dedupedai.com/use-cases/chigwell-penelopa-ai/connect-codex-claude-sessions-to-penelopa/ - Repair broken Penelopa hooks without reinstalling: https://dedupedai.com/use-cases/chigwell-penelopa-ai/repair-broken-penelopa-hooks/ - Trace agent steps back to their evidence: https://dedupedai.com/use-cases/chigwell-penelopa-ai/trace-agent-steps-in-sessions-explorer/ - Capture agent sessions on a machine without desktop: https://dedupedai.com/use-cases/chigwell-penelopa-ai/run-penelopa-hooks-on-headless-machine/ ## How to connect Codex and Claude Code sessions to Penelopa with Penelopa AI Repo: [chigwell/Penelopa.ai](https://github.com/chigwell/Penelopa.ai) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Install Penelopa once so real coding sessions feed its recommendations dashboard. **Who it is for:** A developer who uses Codex or Claude Code every day and wants feedback on repeated workflow habits. **The problem:** Agent sessions run in isolation, so repeated friction and reusable patterns go unnoticed. **The outcome:** Hooks capture real sessions, and the Connection page shows Connected with recommendations available in the dashboard. ### Steps 1. On macOS or Linux, run `curl -fsSL https://penelopa.ai/script | sh` in a terminal. On Windows, use the PowerShell commands from the README. 2. In Codex, open Settings → Hooks, review Stop and SessionEnd, and trust them. In the CLI, use `/hooks`. 3. Open the desktop app and go to Connection to confirm configured agents. 4. Continue a normal coding session so a real hook event is observed. 5. Check that the Connection status changes to Connected, then open the web dashboard for recommendations. ### Prompt to paste into your agent ```text In the chigwell/Penelopa.ai repo, guide me through connecting Codex and Claude Code to Penelopa, then confirm the Connection page shows Connected after one real coding session. Report the status and any hook that still needs trust. ``` ### Caveat The installation self-test uses synthetic data, so Connected only appears after a real agent event is observed and hooks are trusted. Risk, remote script: The install command downloads and runs a shell script from penelopa.ai, so review the script source before running it on a work machine. Risk, credentials: The installer stores an account credential file in your home directory that independent hooks use, and the README says the desktop app does not encrypt it. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/chigwell-penelopa-ai/connect-codex-claude-sessions-to-penelopa/ ## How to repair broken Penelopa hooks without reinstalling with Penelopa AI Repo: [chigwell/Penelopa.ai](https://github.com/chigwell/Penelopa.ai) | Difficulty: Intermediate | Time: 15 minutes | Basis: From the README Run the built-in diagnose and repair commands when capture stops working. **Who it is for:** A developer whose Connection page stopped updating or whose agent hooks seem out of date. **The problem:** Hooks can fall out of date or stop capturing, and reinstalling from scratch is slow and risky to the setup. **The outcome:** Diagnostics show the current state, repair restores Penelopa hooks, and the account token is kept. ### Steps 1. Run `"$(cat "$HOME/.auto-improve/node-path")" "$HOME/.auto-improve/bin/penelopa.cjs" --diagnose` to print diagnostics without credentials or transcripts. 2. Run `"$(cat "$HOME/.auto-improve/node-path")" "$HOME/.auto-improve/bin/penelopa.cjs" --repair` to repair or update the hooks and repeat the local self-test. 3. Open Connection in the desktop app and continue a coding session. 4. Confirm the status returns to Connected after a real event is observed. ### Prompt to paste into your agent ```text In the chigwell/Penelopa.ai repo, run the diagnose command, then the repair command, using the local private runtime path at $HOME/.auto-improve. Report the diagnostic summary and confirm the repair kept the existing account token. ``` ### Caveat Repair keeps the account token unless a new token is supplied, and the README warns that the self-test does not prove an agent has approved its hooks. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/chigwell-penelopa-ai/repair-broken-penelopa-hooks/ ## How to trace agent steps back to their evidence with Penelopa AI Repo: [chigwell/Penelopa.ai](https://github.com/chigwell/Penelopa.ai) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Browse captured sessions and inspect each message and tool input and output. **Who it is for:** A developer or team lead reviewing what a coding agent actually did during a past session. **The problem:** Recommendations are hard to trust without seeing the exact messages and tool calls that produced them. **The outcome:** You can read each captured message and tool call, follow new events, and link process steps to their evidence. ### Steps 1. Open the desktop app and confirm you are connected to an account. 2. Open Sessions from the web dashboard shown in the app. 3. Select a captured conversation and inspect each message and tool input and output. 4. Follow newly delivered events, then trace a process step back to its evidence. ### Prompt to paste into your agent ```text In the chigwell/Penelopa.ai repo, open the Sessions explorer for the session and list the tool calls with their inputs and outputs. Verify that each recommendation you cite links to a message visible in that session. ``` ### Caveat Sessions needs an internet connection, and availability of large content follows server retention. Risk, credentials: Sessions show full captured messages and tool outputs, which may contain secrets from your project, so review what you share from this view. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/chigwell-penelopa-ai/trace-agent-steps-in-sessions-explorer/ ## How to capture agent sessions on a machine without desktop with Penelopa AI Repo: [chigwell/Penelopa.ai](https://github.com/chigwell/Penelopa.ai) | Difficulty: Advanced | Time: 45 minutes | Basis: From the README Install hooks only, with no desktop client, on a server or CI-style machine. **Who it is for:** An engineer who runs Codex or Claude Code on a Linux machine where a desktop app is not wanted. **The problem:** Agent sessions on headless machines are not captured, so their workflows never reach the dashboard. **The outcome:** Hooks capture events on the headless machine, while the dashboard and Sessions explorer are used from a browser. ### Steps 1. Run `curl -fsSL https://penelopa.ai/script | sh -s -- --agent claude --no-desktop` to install Claude Code hooks without building a client, using the README option table. (inferred, check the README) 2. Open the Connection page in the web dashboard and continue a session on the machine. 3. Run `"$(cat "$HOME/.auto-improve/node-path")" "$HOME/.auto-improve/bin/penelopa.cjs" --diagnose` if no event appears. ### Prompt to paste into your agent ```text In the chigwell/Penelopa.ai repo, install hooks for with the --no-desktop option, then run the diagnose command. Report which agents are configured and whether any queued data remains. ``` ### Caveat Linux desktop is not included in this release, so headless hook capture is the only Linux option, and the --agent flag form is inferred from the README option table. Risk, remote script: The install command runs a script downloaded from penelopa.ai on a shared machine, so confirm the script before running it there. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/chigwell-penelopa-ai/run-penelopa-hooks-on-headless-machine/