Borrowing it
Nothing to install: this file belongs to roccodaffuso/sift. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/roccodaffuso/sift/main/.agents/skills/sift-memory/SKILL.mdgit clone --depth 1 https://github.com/roccodaffuso/siftWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/roccodaffuso/sift/sift-memory)<a href="https://agentmods.dev/skills/roccodaffuso/sift/sift-memory"><img src="https://agentmods.dev/badge/skills/roccodaffuso/sift/sift-memory/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/roccodaffuso/sift/sift-memory"><img src="https://agentmods.dev/badge/skills/roccodaffuso/sift/sift-memory.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00075 | $0.00829 |
| Opus 5.5 | $0.00030 | $0.00332 |
| Sonnet 5.5 | $0.00015 | $0.00166 |
| Haiku 4.5 | $0.00007 | $0.00083 |
Grade A, and why
sift-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 2mo 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.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sift Memory
Use sift as a local memory search layer for prior AI coding-assistant sessions. It reads only local Claude Code, Codex, Cursor, and OpenCode logs/databases and writes only its own index under ~/.sift/.
Command Resolution
Prefer the installed CLI:
sift --help
When working inside the sift repository and the global command is unavailable, use:
node ./bin/sift.js --help
Do not use npx or any network-based install path. If neither command works, tell the user to install sift locally.
Actions
sift_index
Refresh the local index before memory-heavy work, after new logs were created, or when search says no index exists:
sift index --json
Use the repo-local fallback when needed:
node ./bin/sift.js index --json
Use a full rebuild only when the user asks, parser/index behavior changed, or results look stale after an incremental run:
sift index --full --json
sift_search
Search prior local sessions with targeted queries:
sift "<query>" --limit 10 --json
Filter by source when useful:
sift "<query>" --tool codex --limit 10 --json
sift "<query>" --tool claude --limit 10 --json
sift "<query>" --tool cursor --limit 10 --json
sift "<query>" --tool opencode --limit 10 --json
Good queries are concrete: project names, command names, file names, bug symptoms, package names, decisions, or user phrasing. Prefer a few focused searches over one broad search.
sift_show
Follow a search result ref when the surrounding conversation is needed:
sift show "<ref>" --json
Use --context N only when the default two messages before and after are insufficient.
sift_list
List recent indexed sessions when you need orientation before choosing search terms:
sift list --limit 10 --json
Filter by source when useful:
sift list --tool codex --limit 10 --json
Workflow
- Use this skill when local session memory could materially improve the answer.
- Run
sift indexunless the task is tiny or the index is known to be fresh. - Run
sift_searchwith 1-3 focused queries and structured JSON output. - Follow only the most relevant result refs with
sift_showwhen more context is needed. - Use the results as private local context. Summarize only the relevant facts needed for the task.
- Do not paste large raw logs. Do not expose secrets, credentials, tokens, private personal data, or unrelated session contents.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2mo ago First seen · 109 lines · 75 tokens per session scan A 798562327742
sift-memory is a skill published in the GitHub repository roccodaffuso/sift (4 stars, last pushed 2mo ago), licensed MIT. It adds 75 tokens to every session and 829 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-08-31.
Other skills, from other repositories
shared-agents-knowledge
Sync and use the shared-agents repo (team skills, rules + learnings). Use at session start, when loading team knowledge, searching learnings, or when the user mentions shared-agents, shared learnings, or team knowledge.
agent-memory
Recall earlier conversations from the user's other AI tools (Claude Code, Claude Desktop, Codex, Cursor, Gemini CLI, Antigravity, Devin), and search the files of the projects they indexed. Use before asking about or assuming a project's history, when the user refers to a prior decision, discussion or fix ("as we…
coordinate-agents
Coordinate a task across multiple AI coding agents (Claude Code, Codex, Gemini, and others). Use when the user asks to delegate work to another agent, get a second opinion or review from a different agent, run agents in parallel on one task, or split a task between agents. The invoking agent becomes the coordinator…
loredex-curate
Write a Start-Here brief for a project or a recent session's findings, and flag stale/duplicate/orphaned notes. Use when the user says "curate my dex/research", "where do I start", "make sense of this project", or just finished a session that produced several markdown files and wants a handoff.
loredex-mcp
Set up or explain the loredex MCP server — live dex access (search, read, handoffs, product state, safe writes) for any MCP-compatible agent. Use when the user wants agents to query the dex mid-task, asks to "connect the dex to Cursor/Codex/an agent", or asks what the loredex MCP tools do.
loredex-init
Set up loredex for the current project — creates/registers the dex and writes AGENTS.md/CLAUDE.md/.cursor rule conventions. Use when the user says "set up loredex", "init loredex", or wants to start tracking this project's research in a dex.