link-memory

A local-memory tool for proposing, saving, finding, reviewing, changing, archiving, restoring, or deleting durable project context through a command-line interface.

In plain words
What is it for?
Use it to recall project conventions, record user-approved decisions, review stored memories, and manage their lifecycle. It proposes memories first unless the user explicitly asks to save one.
Why use it?
Important decisions and preferences can otherwise be lost between work sessions. It keeps approved context available without requiring an MCP connection.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/gowtham0992/link/link-memory
Any agent
npx skills add gowtham0992/link --skill link-memory
Clone the repo
git clone --depth 1 https://github.com/gowtham0992/link

Made for: Claude Code, Codex.

Per session 52 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 601 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00052 $0.00601
Opus 5 $0.00026 $0.00300
Sonnet 5 $0.00010 $0.00120
Haiku 4.5 $0.00005 $0.00060

Measured 2d ago against content hash 3aa8ee93bd60, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

link-memory scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/link-memory/SKILL.md · 51 lines

What it actually says

Use this skill after important user-approved decisions, preference changes, project conventions, or long work sessions that may deserve durable context. In a source checkout, replace lnk with python3 link.py. Do not silently save durable memory; propose first unless the user directly asks to remember, approves a proposal, or explicitly confirms an important decision should become durable memory.

If Link session hooks are installed for this agent, the memory brief is injected automatically at session start — skip step 1 and go straight to task-specific recall.

  1. Prime before work:

    lnk brief "<current task>" [link-root]
    
  2. Recall specific memory:

    lnk recall "<topic>" [link-root]
    
  3. End a session with proposal-only memory candidates:

    lnk session-end <session-notes-or-transcript> [link-root] --limit 3
    

    Use - as the input when piping a transcript on stdin. Show the proposals to the user; do not save durable memory until the user approves one.

  4. Save an explicit memory:

    lnk remember "<user-approved memory>" [link-root] --type note --scope user
    

    Use --project <slug> for project-scoped memory, --visibility private|project|team for sharing intent, --review-after YYYY-MM-DD for stale-risk memories, and --expires-at YYYY-MM-DD for temporary context. When a brief or recall reports a memory backlog (pending captures or reviews above threshold), offer the user a short consolidation pass:

    lnk consolidate [link-root]
    

    The plan is read-only: it groups duplicates and recurring themes and prints accept/discard/review commands. Apply an action only after the user approves it.

  5. Review and explain before trusting uncertain memory:

    lnk memory-inbox [link-root]
    lnk explain-memory <name-or-title> [link-root]
    lnk review-memory <name-or-title> [link-root]
    
  6. Change lifecycle safely:

    lnk update-memory <name-or-title> "<new text>" [link-root]
    lnk archive-memory <name-or-title> [link-root] --reason "<why>"
    lnk restore-memory <name-or-title> [link-root]
    lnk forget-memory <name-or-title> [link-root] --confirm
    

When duplicate or conflict warnings appear, prefer updating, reviewing, or archiving existing memory over creating another page.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 51 lines · 0 tokens per session scan A 3aa8ee93bd60

Subscribe to this mod's changes

link-memory is a skill published in the GitHub repository gowtham0992/link (180 stars, last pushed 17d ago), licensed MIT. It adds 52 tokens to every session and 601 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

llm-wiki

Build and maintain a persistent, interlinked Obsidian-compatible markdown wiki using Karpathy's LLM Wiki pattern. Extension-backed with auto-generated metadata, guardrails, and 14 custom tools (+3 opt-in agent-trajectory tools).

zosmaai/pi-llm-wiki · 52 tokens

youtube-fetcher

Turn a YouTube video into a structured, Obsidian-ready archival Markdown note containing its transcript, creator metadata, description, chapters, language, and capture provenance. Use when a user shares a YouTube URL or video ID and wants transcripts, captions, subtitles, notes, a knowledge-base record, summarization…

JimmySadek/youtube-fetcher-to-markdown · 102 tokens

ingest

Turn messy source notes into structured wiki pages (frontmatter, wikilinks, one folder per type) and update the vault's index, log and hot cache. Use when the user drops files into inbox/ and asks to process them, points at a note or URL to ingest, or pastes raw content to file. Triggers on: "ingest", "ingesta"…

sturlese/hippocampus · 110 tokens

lint

Health check for the vault: runs the deterministic linter (frontmatter, dead links, orphans, duplicates, index drift) plus editorial checks (stale claims, unlinked mentions, missing pages), then fixes what the user approves. Triggers on: "lint", "revisa el vault", "health check", "chequea el wiki", "limpia el vault"…

sturlese/hippocampus · 84 tokens

save

File the current conversation's valuable content (analysis, decision, session summary) as a structured wiki note so it isn't lost in chat history. Triggers on: "save this", "guarda esto", "guárdalo en el wiki", "/save", "$save", "file this", "apunta esto", "guarda esta conversación", "save this session".

sturlese/hippocampus · 78 tokens

obsidian-setup-consultant

Use this skill whenever the user wants to install, configure, migrate, design, or connect Obsidian as a second brain, knowledge base, cloud-synced vault, or AI/Hermes-readable memory. It is especially important when the user asks for an Obsidian setup prompt, vault architecture, folder structure, plugin…

SikJa/hermes-obsidian-setup-consultant · 124 tokens