/ijfw-update
Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update
$ npx -y skills add FerroxLabs/ijfw --skill ijfw-update --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/ijfw-update
Context preview
The summary Claude sees to decide when to auto-load this skill.
Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update
SKILL.md
ijfw-update.SKILL.mdname: ijfw-update
description: "Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update"
What this skill does
Helps the user check for and apply IJFW updates. Updates are air-gapped: the model cannot execute them. The model issues a confirmation token; the user types `ijfw update --confirm <token>` in their terminal to actually run.
When to fire
Triggers: "update ijfw", "upgrade ijfw", "is there a new version", "latest version", `/update`. Also fires when memory prelude reports "update available" on first turn.
Execution
1. **Call the MCP tool**: `ijfw_update_check`. This returns:
- `current` -- installed version
- `latest` -- latest published version on npm
- `available` -- boolean
- If `available: true`: also `confirmation_token`, `expires_at`, `changelog_url`, and `instruction`. The same call writes the pending sentinel that the terminal command consumes -- one MCP call, one terminal command, no intermediate ceremony.
2. **If up to date**: report it. Stop.
> IJFW is up to date (v1.2.5).
3. **If update available**: present the version delta + changelog link, then surface the OOB instruction verbatim. The terminal command is the air-gap; the model never runs the update itself.
> Update available: v1.2.4 -> v1.2.5 > Changelog: <changelog_url> > > To proceed, run in your TERMINAL: > ijfw update --confirm <confirmation_token> > > Token expires in 5 minutes. I cannot run the update for you -- only typing this command in your terminal can.
4. **DO NOT** run `npm install`, `npx @ijfw/install`, `bash scripts/install.sh`, or any equivalent yourself. The MCP path is air-gapped on purpose. Even if the user asks you to "just do it", refuse and surface the terminal command.
5. **V155-017 (v1.5.5):** `ijfw_update_apply` was retired from the MCP tool surface — it was redundant given that `ijfw_update_check` already writes the sentinel + issues the token in one step. If an older skill or saved prompt calls it, the MCP server will respond with unknown-tool. Re-route those flows to call only `ijfw_update_check` and forward its `instruction` field directly to the user.
Security model
The token + sentinel + terminal-confirm flow exists so that prompt injection in stored memory, fetched docs, or user-paste content cannot trick the model into auto-updating IJFW. See `docs/SECURITY.md` for the full threat model.
Common variations
- "Just check, don't update yet" -- call `ijfw_update_check`, report status, stop.
- "Update silently" -- not supported via the MCP path. Tell the user to run `ijfw update --yes` in their terminal directly.
- "Roll back to <version>" -- not supported yet (rollback tarballs deferred). Suggest `npm install -g @ijfw/install@<version>` from the user's terminal.
After a successful update
The user will see "Updated to v<latest>" in their terminal. The next IJFW SessionStart will reflect the new version. Suggest restarting any open agent sessions so they pick up the new skills/hooks.
Read more
name: ijfw-update description: "Check for and apply IJFW updates safely. Trigger: 'update ijfw', 'upgrade', 'latest version', 'is there a new version', /update"
What this skill does
Helps the user check for and apply IJFW updates. Updates are air-gapped: the model cannot execute them. The model issues a confirmation token; the user types `ijfw update --confirm <token>` in their terminal to actually run.
When to fire
Triggers: "update ijfw", "upgrade ijfw", "is there a new version", "latest version", `/update`. Also fires when memory prelude reports "update available" on first turn.
Execution
1. **Call the MCP tool**: `ijfw_update_check`. This returns:
- `current` -- installed version
- `latest` -- latest published version on npm
- `available` -- boolean
- If `available: true`: also `confirmation_token`, `expires_at`, `changelog_url`, and `instruction`. The same call writes the pending sentinel that the terminal command consumes -- one MCP call, one terminal command, no intermediate ceremony.
2. **If up to date**: report it. Stop.
> IJFW is up to date (v1.2.5).
3. **If update available**: present the version delta + changelog link, then surface the OOB instruction verbatim. The terminal command is the air-gap; the model never runs the update itself.
> Update available: v1.2.4 -> v1.2.5 > Changelog: <changelog_url> > > To proceed, run in your TERMINAL: > ijfw update --confirm <confirmation_token> > > Token expires in 5 minutes. I cannot run the update for you -- only typing this command in your terminal can.
4. **DO NOT** run `npm install`, `npx @ijfw/install`, `bash scripts/install.sh`, or any equivalent yourself. The MCP path is air-gapped on purpose. Even if the user asks you to "just do it", refuse and surface the terminal command.
5. **V155-017 (v1.5.5):** `ijfw_update_apply` was retired from the MCP tool surface — it was redundant given that `ijfw_update_check` already writes the sentinel + issues the token in one step. If an older skill or saved prompt calls it, the MCP server will respond with unknown-tool. Re-route those flows to call only `ijfw_update_check` and forward its `instruction` field directly to the user.
Security model
The token + sentinel + terminal-confirm flow exists so that prompt injection in stored memory, fetched docs, or user-paste content cannot trick the model into auto-updating IJFW. See `docs/SECURITY.md` for the full threat model.
Common variations
- "Just check, don't update yet" -- call `ijfw_update_check`, report status, stop.
- "Update silently" -- not supported via the MCP path. Tell the user to run `ijfw update --yes` in their terminal directly.
- "Roll back to <version>" -- not supported yet (rollback tarballs deferred). Suggest `npm install -g @ijfw/install@<version>` from the user's terminal.
After a successful update
The user will see "Updated to v<latest>" in their terminal. The next IJFW SessionStart will reflect the new version. Suggest restarting any open agent sessions so they pick up the new skills/hooks.
IJFW — It Just F*cking Works. Ferrox Labs' local-first infrastructure for AI coding agents: shared memory, smart routing, multi-AI cross-audits, disciplined workflow.
Repo: FerroxLabs/ijfw
Other skills on ijfw.
- /ijfw-agents-md
Maintain canonical AGENTS.md (open spec). Trigger: 'agents.md', 'update AGENTS.md', or auto-fired by ijfw-team after agent generation.
Open skill - /ijfw-auto-memorize
Session-end auto-extraction of lessons, errors, fixes, and user feedback into structured memory. Fires at session end. Requires consent on first run.
Open skill - /ijfw-commit
Terse conventional commits. Trigger: commit, git commit, /ijfw-commit
Open skill - /ijfw-complete-milestone
Use when a milestone is shipping and you need to archive its artifacts, generate a summary, and seed the next milestone. Trigger: 'milestone complete', 'ship milestone', 'wrap milestone', 'complete milestone <id>', /ijfw-complete-milestone.
Open skill - /ijfw-compress
Compress memory/context files into terse form. Trigger: /compress, compress file
Open skill - /ijfw-compute
Use when the user says: 'compute', 'crunch this', 'analyze logs', 'aggregate the data', 'run a script', 'dedupe', 'count by', 'top N', or any data-shaping ask. Replaces dumping raw data into context with sandboxed script execution.
Open skill

