claude

fpl-hsmem

Project-scoped cross-session memory for Hindsight API >=0.10.2 and <0.11.0. Use when a repository explicitly opts in via .hindsight.json; hooks default off, enrichment is operator-gated, and BankGate is client safety, not server authorization.

4 on GitHubby ForgePlanv3.7.3MITupdated 12 days agomemory
/

Plugin fpl-hsmem from marketplace ForgePlan-marketplace in ForgePlan/marketplace. Commands read from the repo's manifest — 12 days ago.

This plugin runs code on your machine — its hooks fire automatically on events like tool use and it starts local MCP servers. The exact commands are listed below; read them before installing.

What it installs

Skills (10)

  • audit-bank

    Read-only audit of a bank in six calls, ending in a written verdict rather than a number. Answers the three questions a count cannot: is it safe to keep filling, is it actually working, and did anyone choose this bank on purpose. EN: Posture audit — masking on or off, what a single document is worth

  • bootstrap

    One-time seeding of a bank for a project that has just been activated: set the bank's persona, ingest the formal artifacts that already exist, and create a small starter set of knowledge pages. Not the technical wiring — that is the setup CLI — but the layer that decides what this bank is for before

  • correct-memory

    The workflow that makes a bank correctable instead of cumulative: find the exact stored row, retire it with a stated reason, write the correction in full, rebuild the beliefs that rested on it, and verify the background job finished. Deleting a wrong fact is the easy answer and the wrong one — six m

  • diagnose

    The full diagnostic when memory is not doing what it should: server reachable, bank resolved as intended, hooks actually firing, config not overridden somewhere you forgot, and jobs completing. Written to distinguish the failure modes that look identical from the outside — a dead server and an empty

  • directives

    The standing rules a bank follows when it synthesises an answer — precedence between conflicting decisions, what an answer must always name, what to leave out. The distinction that makes this useful: a factually wrong answer is a memory problem, a consistently badly-shaped answer is a rules problem,

  • enrich

    Turns knowledge the team already has — design records (ADRs, RFCs, notes, problem and evidence records), decision ledgers, review threads, chat threads — into short, self-contained memory items (a decision and why, a rejected option and why, a lesson, a pitfall, a rule) that link back to the artifac

  • export-bank

    A readable markdown snapshot of a bank for backup, audit or handing context to a person. Says plainly what it is not: this is for humans, and moving memory between banks is a different mechanism entirely. EN: Snapshot a bank to disk before something risky, or to share context without giving out serv

  • memory-setup

    Look at a project and propose the retrieval and memory layers it should have — what to install, what each layer answers, and what it will never answer. The reason this exists: an agent that can only grep is slow in a way nobody notices, and an agent with no memory re-learns the same lesson every ses

  • mental-model

    Guided creation of a living knowledge page — a standing answer to a recurring question, rebuilt from memories after every consolidation. The whole value is in the source query: a narrow question produces a page worth reading for months, "everything about X" produces noise that nobody opens. EN: Crea

  • status

    A thirty-second read of one memory bank: which bank, who chose it, how much is in it, and whether anything is silently broken. The first thing to run when memory "feels off" — most such feelings turn out to be the session reading one bank while something else writes to another. EN: Quick health + st

Subagents (1)

  • memory-curator

    Methodology: curation of a Hindsight memory bank through this relay — the correction cycle (find → retire with a reason → write the correction → rebuild what rested on it → verify the job finished) plus the read discipline that decides between searching, enumerating and synthesising. Scoped to MEMOR

MCP servers (1)

  • hindsight

    runs: node ${CLAUDE_PLUGIN_ROOT}/dist/index.mjs

Hooks (3)

  • UserPromptSubmit

    node

  • Stop

    node

  • SessionEnd

    node

View source on GitHub forgeplan.dev

More from ForgePlan/marketplace

All 24 plugins →

Flagship workflow plugin of the ForgePlan ecosystem — the skills, templates and authoring contract the rest of the marketplace is written against: the smith methodology router (14 contexts), the agent-authoring guide every forgeplan-aware agent follows, and the gate, cleanup and insight skills that keep a pipeline honest. Use when you want the methodology itself rather than one specialist — deciding how to approach a task, gating an artifact before activation, or authoring a new agent to the house canon. Full history in CHANGELOG.md.

44 skills · 1 agent · hooks

4

Professional specialist agents: security and prompt-injection review, PII detection, architecture and staff-level system review, the pre-activation guardian gate, ADR and specification authors, research analysts and infrastructure roles. Use when a task needs a named expert rather than a generalist — an OWASP/STRIDE pass, an architectural fitness review, a decision recorded as an ADR, or the last gate before an artifact is activated. Full history in CHANGELOG.md.

30 agents

4

CANVAS design-system-to-code methodology: a coordinator walks a Pencil design through capture, audit, norm-check, vectorize and assemble, dispatching every phase and every verifier as a separate isolated context with a blocking generator-is-not-verifier gate at each handoff, plus a fail-closed hook that blocks source writes until the token contract is active. Use when a design system has to become working components in the project's own framework, with visual parity and token fidelity proven rather than asserted. Full history in CHANGELOG.md.

6 skills · 4 commands · 8 agents · hooks

4

Builds a navigable map of a project — zones, nodes and verified edges — as a `map.json`, through an eight-agent pipeline (scan -> extract -> verify -> emit -> validate) where every stage runs in a fresh isolated context and a deterministic guardian script, not an LLM, is the sole authority on whether the result can be trusted. Use when you need a structural view of a large or unfamiliar repository that a file tree cannot give, or when a project's forgeplan artifact graph should be shown alongside its code. Not a documentation generator: an edge is either sourced from the artifact graph or proven by a grep, and an unproven one is dropped rather than shipped as noise. Full history in CHANGELOG.md.

3 skills · 4 commands · 8 agents · hooks

4

Brownfield extraction pack — turns legacy code and documents into a structured forgeplan graph through a discover agent and twelve extraction skills (ubiquitous language, use cases, invariants, causal links, scenarios, knowledge-graph curation, reproducibility). Use when taking over an undocumented codebase and the first problem is not what to build but what is already there and why it is that way. Full history in CHANGELOG.md.

12 skills

4

Core engineering agents: the source-mutating coder, the code reviewer and test runner that record evidence, plus debugger, planner, researcher, performance and production-readiness roles. Use when the work is ordinary engineering — implement this, review that diff, run the suite and report coverage against the acceptance criteria — and the reviewer should be a different context from the author. Full history in CHANGELOG.md.

11 agents

4

Similar plugins

claude-mem

claudecodex

Persistent memory for Claude Code: the next session starts where the last one left off.

29 skills · 1 MCP server · hooks

thedotmack/claude-mem 116,618

Tenant-isolated AI teams that plan work, coordinate tasks, recall approved context, and return evidence-backed results

6 skills · 6 commands · 4 agents · 1 MCP server

ruvnet/ruflo 81,589

mem0

claudecodex

Cross-session memory and token savings for coding agents.

6 skills · 1 agent · 1 MCP server · hooks

mem0ai/mem0 68,893

mempalace

claudecodex

AI memory system — mine projects and conversations into a searchable palace. 47 MCP tools, auto-save hooks, guided setup.

3 skills · 6 commands · 1 MCP server · hooks

MemPalace/mempalace 60,536

DuckDB-powered skills for Claude Code: read any data file, attach and query DuckDB databases, search DuckDB/DuckLake docs, search past session logs, and install/update DuckDB extensions.

9 skills

anthropics/claude-plugins-official 47,076

pensyve

claude

Universal memory runtime for Claude Code with cross-session memory, entity-aware recall, lifecycle hooks, skills, commands, and memory-curator agents

7 skills · 7 commands · 2 agents · hooks

wshobson/agents 42,461

Maintain fpl-hsmem?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins24 plugins