Review an agent’s saved memory alongside project and personal instructions. Receive proposed destinations for useful entries and a cleanup report covering duplicates, outdated facts, conflicts, and unclear items, with approval required before changes.
Review auto-memory entries and propose promotions to OPENSPACE.md, OPENSPACE.local.md, or retained auto-memory topics; detect duplicates, stale facts, conflicts, and ambiguous entries before making changes.
Review the instructions the agent will receive before you install.
Review the user's memory landscape and produce a clear report of proposed changes, grouped by action type. Do not apply changes until the user approves them.
Read OPENSPACE.md and OPENSPACE.local.md from the project root if they exist. Review auto-memory from the memory section and MEMORY.md; use memory_read for topic files when needed. Note team memory only as unavailable unless this deployment explicitly has a team memory backend.
Success criteria: you have the contents of all available memory layers and can compare them.
For each substantive auto-memory entry, determine the best destination:
| Destination | What belongs there | Examples |
|---|---|---|
OPENSPACE.md | Project conventions and instructions for OpenSpace that all contributors should follow | use uv not raw pip; API routes use kebab-case; test command is pytest; prefer functional style |
OPENSPACE.local.md | Personal instructions specific to this user or machine, not applicable to other contributors | concise responses; explain tradeoffs; do not auto-commit; run tests before committing |
| Auto-memory topic | Durable cross-session user, feedback, project, or reference memory that does not belong in static project instructions | user role, durable feedback, external reference pointers, uncertain but useful context |
| Stay put | Working notes, temporary context, or entries that do not clearly fit elsewhere | session-specific observations, uncertain patterns |
Important distinctions:
OPENSPACE.md and OPENSPACE.local.md contain instructions for the agent, not arbitrary external-tool preferences.Success criteria: each entry has a proposed destination or is flagged as ambiguous.
Scan across all layers for:
OPENSPACE.md or OPENSPACE.local.md.MEMORY.md entries that contain content that belongs in a topic file.Success criteria: all cross-layer issues are identified.
Output a structured report grouped by action type:
If auto-memory is empty, say so and offer to review OPENSPACE.md for cleanup.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Prepare audio stems, the separate parts of a track, for mixing by matching their durations with silence, loops, or crossfades. Verify each generated and processed file before moving to the next production step.
Produce audio in separate stems, meaning individual instrument tracks, and verify each output as you go. Align track lengths with silence, loops, or crossfades, then apply effects, mix a master, and package the files.
Build serverless endpoints that relay external API requests for a dashboard while keeping API keys on the server. Includes Vercel examples for input checks, caching, cross-origin access, and consistent JSON responses.
Produce a master track and separate audio parts through a structured Python workflow. Apply filters and reverb, export WAV files with consistent settings, package them in a ZIP archive, and check that the deliverables meet the requested specifications.
Keep document creation moving when web searches and page retrieval repeatedly fail. The skill preserves the original task, uses the model’s existing knowledge, and clearly identifies information that could not be checked against external sources.
When Python execution repeatedly fails in execute_code_sandbox, use a file-based alternative: save the script, run it through the shell, and capture results or errors. The guidance covers paths, imports, timeouts, structured output, and optional cleanup.