agentsclimarketplace

Memory

Skill Neznakometz/StackForge/core/skills/memory

Thin core + domain sets for a fast, stack-aware project start. Cross-agent (Claude Code, Cursor, Codex, Gemini) via the SKILL.md standard. MIT.

Install
npx -y skills add Neznakometz/StackForge --skill memory

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

The agent's memory between sessions — context management (L1) and checkpoints (L2). Apply when the context fills up, when moving to a new chat, and when closing a session.

SKILL.md

1.7 KB, 403 tokens by cl100k_base, as published. Nobody here has run it

Memory — L1 (context) + L2 (checkpoints)

L3 (durable/SQLite) and the rolling log (Layer 0.5) are in a separate set, storage (/plugin install storage@stackforge), installed when persistent memory is needed. Here — L1+L2.

L1 — active context

  • Watch the fill level of the context window. At 40% — propose /compact. Never let it reach auto-compact at 95% (quality degrades).
  • /compact: a brief summary of progress (task, done, remaining) → continue.
  • /convolife: estimate the % of context filled, report what's left.
  • /clear: a full wipe — only when the topic is fully closed.

L2 — checkpoints

  • /checkpoint — before every chat close. Creates Sessions/checkpoints/YYYY-MM-DD-checkpoint.md: ## What was done (specifics: paths, numbers) · ## Key decisions (why this way) · ## Not finished (the next step).
  • Auto-checkpoint: a task spanning 3+ files or 10+ steps → write progress to Sessions/tmp/progress-<slug>.md every exactly 10 tool calls; delete it on completion.
  • /newchat: checkpoint → "start a new session".

Injection order for a new session (top to bottom)

[Memory] (key facts) → [Previous session] (the last checkpoint) → the user's message.

Rules

  • Temporary (task details) → only L2, not into long-term memory.
  • Permanent (patterns, paths, architecture) → L3, if enabled; otherwise into agent-memory.md.
  • Replace stale entries, don't accumulate contradictions.
  • PROGRESS.md is maintained by the orchestrator (task-loop), the checkpoint — more detailed, for recovery.

Gives 0 of the 12 instructions most memory context skills give in 403 tokens

Counted across 674 of the 847 authors here whose files we hold, read 2026-08-06

  • inform the user when setup is completein 21 of 674, across 6 files
  • confirm the draft with the user before writingin 21 of 674, across 6 files
  • update the agent skills block in place if it existsin 21 of 674, across 6 files
  • present findings to the userin 20 of 674, across 5 files
  • write the three docs files from seed templatesin 20 of 674, across 5 files
  • ask the user about each decision one at a timein 19 of 674, across 4 files
  • edit CLAUDE.md if it existsin 18 of 674, across 3 files
  • explore current repo statein 18 of 674, across 3 files
  • do not overwrite user edits to surrounding sectionsin 18 of 674, across 3 files
  • back up the original file before overwritingin 16 of 674, across 8 files
  • keep the memory index under 200 linesin 15 of 674
  • Provide actionable steps and verificationin 13 of 674, across 2 files

Said here and by no other author read

  • propose compact at 40% context fill
  • prevent context from reaching auto-compact
  • write a checkpoint before closing chat
  • auto-checkpoint long tasks every 10 tool calls
  • delete progress file upon task completion
  • store temporary task details only in L2

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.