agentleFS
shashu10shashu10/personal-finance-skills/skills/finance-memory/SKILL.md

finance-memory

Skill2 starsChanged 51 days agoMITView on GitHub

Save and retrieve confirmed personal financial facts, decisions, and their corrections in a private workspace with provenance and supersession. Use when the user provides an update, confirms a choice, or asks to resume prior financial planning.

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. About 0 are common, 8 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use the installed finance-core in its dedicated environment and selected private workspace.

    line 14
  2. Very specific:

    Read the selected private workspace's setup.md and the active project's applicable AGENTS.md, when present, to find the interpreter, record …

    line 16
  3. Very specific:

    Use the selected normalized workspace without overwriting a project's original records.

    line 18
  4. Very specific:

    Run the commands yourself.

    line 19
  5. Very specific:

    Read facts.json, decisions.json, and relevant dated reports.

    line 20
  6. Distinctive:

    Read memory records for exact fields and examples.

    line 23
  7. Distinctive:

    Keep personal records, source notes, and input envelopes outside the public checkout.

    line 24
  8. Distinctive:

    Do not store credentials, authentication cookies, recovery codes, or unnecessary personal identifiers.

    line 25
  9. Distinctive:

    If a user shares a fact during an authorized intake or asks to remember it, save their statement without adding assumptions or requesting …

    line 40
  10. Distinctive:

    Ask only when the status, meaning, or material conflict cannot be resolved from context.

    line 42
  11. Very specific:

    Use stable IDs. Repeating identical records is idempotent.

    line 46
  12. Very specific:

    Run finance-core validate --workspace /private/path and inspect the saved records.

    line 53
  13. Distinctive:

    Prefer the latest applicable confirmed fact or decision.

    line 63
  14. Distinctive:

    If the user asks for deletion, inspect the affected records and references and follow the authorized removal request.

    line 72
  15. Distinctive:

    Do not announce reminders, scheduled reviews, broker changes or future automatic updates unless they were separately requested and actually …

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Finance memory

Preserve what the user established and decided, while keeping suggestions and unknowns distinct. Use these files to carry facts and decisions between sessions; do not claim hidden cross-chat memory or automatic session hooks.

Read before writing

Use the installed finance-core in its dedicated environment and selected private workspace. Run finance-core --help if needed; this skill requires no sibling skill. Read the selected private workspace's setup.md and the active project's applicable AGENTS.md, when present, to find the interpreter, record locations, and project adapters. Use the selected normalized workspace without overwriting a project's original records. Run the commands yourself; explain results and needed inputs in ordinary language. Read facts.json, decisions.json, and relevant dated reports. Resolve supersedes chains and inspect conflicts before answering from memory.

Read memory records for exact fields and examples. Keep personal records, source notes, and input envelopes outside the public checkout. Do not store credentials, authentication cookies, recovery codes, or unnecessary personal identifiers. A private alias is usually enough.

Classify the update

  • A fact is a user-confirmed statement with a date and source. Preserve whether the user reported it directly or confirmed a document's contents.
  • A proposal is an idea or recommendation. It is not an accepted decision.
  • A confirmed decision requires explicit user acceptance, including acceptance already present in the conversation; do not ask for it again.
  • An executed decision requires evidence of completion or an explicit user report of completion. Preserve that source; a requested action or prepared order is not a fill.

Tax interpretations, model forecasts, and guessed preferences are not personal facts. Save them in a dated research/scenario report or as an unconfirmed proposal. If a user shares a fact during an authorized intake or asks to remember it, save their statement without adding assumptions or requesting duplicate confirmation. Ask only when the status, meaning, or material conflict cannot be resolved from context.

Save without rewriting history

Use stable IDs. Repeating identical records is idempotent; a changed record with the same ID is a conflict. Correct a record with a new ID and supersedes pointing to the old one. Explain the correction and effective date, and retain its source. For changed decisions, distinguish a new preference from correction of an old error.

Prepare a private envelope containing facts and/or decisions, then run: finance-core memory --workspace /private/path --input /private/path/update.json. Run finance-core validate --workspace /private/path and inspect the saved records. The confirmation field records supplied evidence; the CLI cannot independently prove user consent or a transaction. Do not set it true to bypass a validation error.

Memory changes do not update balances or execute choices. Use sourced account imports for positions and cash. A newly reported purchase can be saved as a user-reported execution while the ledger remains awaiting reconciliation; state that distinction.

Use memory in later work

Prefer the latest applicable confirmed fact or decision. Explain contradictions instead of inventing an explanation for them. Preserve dates and conditional choices, such as a plan that applies only after a job starts or a debt is repaid. Review facts whose circumstances or rules may have changed before using them again.

Model-sharing rules govern exports, not memory storage. Saving a sensitive fact does not authorize uploading the file to another service or a separate model API. Reading it in a hosted agent can itself send its contents to that agent's provider; the export switches do not restrict the host's filesystem or computer-use tools. If the user asks for deletion, inspect the affected records and references and follow the authorized removal request. Append-only correction does not prohibit deleting the user's private data.

Report the result

Summarize what was saved, superseded, left unresolved, or kept as a proposal. Link the actual private files and state any remaining reconciliation need. Do not announce reminders, scheduled reviews, broker changes or future automatic updates unless they were separately requested and actually configured.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.