DOCUMENTATION TOOLS / SESSION REVIEW
Reflect / Daily Reflection
I built Daily Reflection to have AI review my recorded work across Copilot sessions and improve the instructions used for the next session. The workflow automatically processed the session records, identified corrections and recurring problems, saved its findings, and fed relevant information into my weekly reports.
MY CONTRIBUTION
I designed and connected cross-session review, transcript processing, instruction updates, and reporting. AI carried out the review and recordkeeping; I used AI coding assistance to develop the tools and reviewed changes that needed my judgment.
TOOL TYPE
An AI reflection workflow using automatically saved Copilot sessions, a Node.js transcript parser, structured summaries, instruction files, and shared reporting records.
THE WORK
Reviewing more than the open conversation
My work with an assistant was spread across sessions. I wanted the workflow to review those records for corrections, failed commands, repeated problems, and changes that worked.
Copilot saved the session records automatically. I added transcript processing so the assistant could select and parse those records, extracting requests, skills used, tool calls, edited files, and failures. AI then compared the structured activity with the existing instructions and identified useful improvements.
WORKFLOW
How AI reviewed the work and carried it forward
The assistant captured explicit corrections during a conversation. When reflection ran, it selected records since the previous review, processed the transcripts, and looked for actionable findings. It saved the reflection and processing timestamp so later runs could pick up from the recorded point.
- 01AI + parserProcess sessionsSelect saved sessions and extract the structured activity.
- 02AIAssess findingsCompare corrections and recurring problems with existing instructions.
- 03AI + NickUpdate and reviewApply routine additions; bring workflow changes to me for review.
- 04AISave and routeLog the reflection, update the processing record, and add reporting notes.
The assistant handled collection, analysis, recordkeeping, and routine edits. Changes to workflow logic retained a review step.
AI appended accomplishments, lessons, and tooling changes to the weekly notes. It also logged skill-specific problems separately; the reporting workflow read that log when it generated the next report.
EXAMPLE
An outdated command in a workflow
Illustrative reconstruction · fictional command issue · not an original session
| Recorded activity | AI reflection follow-up |
|---|---|
| A command from the instructions fails because its option has changed. | Identify the instruction that supplied the old option. |
| The corrected command works after the option is checked. | Propose updating the command reference, keeping the correction tied to the observed issue. |
| The instruction update is completed. | Add the tooling change and lesson to the weekly notes. |
The resulting instruction change affects future work that reads that file. It is not model training or a guarantee that an assistant will permanently remember the correction.
DEVELOPMENT & USE
What the development records contain
- The request for cross-session reviewA preserved Copilot session records my request to review across sessions, an investigation of the available transcript records, and a subsequent reflection request.
- The parser and workflow changeMy development history adds the transcript parser and transcript-based Daily Reflection instructions.
- Reflection in the weekly notesSaved notes contain reflection additions and record the connection between reflection, the tool-problem log, and weekly reporting.
These records establish development and use of the workflow. Some session records show only a request and initial checks, so they should not be read as proof that every requested run completed.
This page summarizes the relevant evidence without reproducing private session content or weekly work records.
RELATIONSHIPS & CREDIT
Where source learning and reporting fit
Daily Reflection existed before I used Learn from Source to review Tamir Dresher’s Squad Skills. That review contributed the pattern for capturing explicit corrections during a session. The reflection instructions credit Dresher’s Reflect plugin for that addition.
The tools shared the records produced by the workflow. AI captured completed work as it happened, reflection added lessons and tooling problems, and Weekly Reports organized those inputs into the weekly update.
The archived implementation used the session formats and tools available in my work environment. Its improvements were saved in instructions and working records that the assistant could use in later sessions.