hooks/webfetch-cache.ts. Options: ttlSeconds, maxEntries (set in user or managed settings pluginConfigs, or /config). Run claude plugin validate on the installed folder to see every event it hooks and every $ call it makes.
webfetch-cache
Short-circuits repeated WebFetch calls for the same URL + prompt within a
session. On a cache hit the hook answers { result } itself without calling
next (no network call); on a miss it awaits the real fetch, stores the
tool's record and returns what came back.
Only the tool's own record (result) is cached, never core's ref: that
number names the messages core produced for one specific call and must not
be replayed on another.
A long page read in pieces (WebFetch's offset, Claude Code 2.1.290+) is cached
piece by piece: a read further down the page never gets the first page back.
Options
ttlSeconds: number how long an entry stays fresh (default 900)
maxEntries: number cache size (default 200)Declared in .claude-plugin/plugin.json (userConfig). Set them in /config, in user settings (~/.claude/settings.json, not project settings), with --settings <file> or in managed settings:
{ "pluginConfigs": { "webfetch-cache@skills-dir": { "options": { } } } }Install
npx claude-code-templates@latest --mod productivity/webfetch-cache
claudeIt is written to .claude/skills/webfetch-cache/, which Claude Code auto-loads as webfetch-cache@skills-dir. For one session with hot reload: claude --plugin-dir .claude/skills/webfetch-cache. claude plugin validate .claude/skills/webfetch-cache prints every event it hooks and every $ call it makes.
Requirements. Mods are on by default in Claude Code 2.1.287+. Typed against Anthropic's declarations: https://github.com/anthropics/claude-code/tree/main/mods
webfetch-cache
Short-circuits repeated WebFetch calls for the same URL + prompt within a
session. On a cache hit the hook answers { result } itself without calling
next (no network call); on a miss it awaits the real fetch, stores the
tool's record and returns what came back.
Only the tool's own record (result) is cached, never core's ref: that
number names the messages core produced for one specific call and must not
be replayed on another.
A long page read in pieces (WebFetch's offset, Claude Code 2.1.290+) is cached
piece by piece: a read further down the page never gets the first page back.
Options
ttlSeconds: number how long an entry stays fresh (default 900)
maxEntries: number cache size (default 200)Declared in .claude-plugin/plugin.json (userConfig). Set them in /config, in user settings (~/.claude/settings.json, not project settings), with --settings <file> or in managed settings:
{ "pluginConfigs": { "webfetch-cache@skills-dir": { "options": { } } } }Install
npx claude-code-templates@latest --mod productivity/webfetch-cache
claudeIt is written to .claude/skills/webfetch-cache/, which Claude Code auto-loads as webfetch-cache@skills-dir. For one session with hot reload: claude --plugin-dir .claude/skills/webfetch-cache. claude plugin validate .claude/skills/webfetch-cache prints every event it hooks and every $ call it makes.
Requirements. Mods are on by default in Claude Code 2.1.287+. Typed against Anthropic's declarations: https://github.com/anthropics/claude-code/tree/main/mods
Scanned by
NVIDIA SkillSpector
Automated static analysis (no LLM) · v2.12.0 · 2026-11-19
Partial analysis: some checks could not run on this component.
4 observations in 3 categories
- Agent Snooping 2
- MCP Least Privilege 1
- MCP Rug Pull 1
SkillSpector looks for patterns such as shell commands, credential access, network calls or unpinned dependencies. An observation is something worth reading in the files before you install, not a confirmed vulnerability. Components that run code locally, like hooks, MCP servers and security tooling, naturally show more of them. Open one to see the lines it points at.
Agent Config Directory AccessAgent Snooping · README.md:23
| 21 | ``` |
| 22 | |
| 23 | Declared in `.claude-plugin/plugin.json` (`userConfig`). Set them in `/config`, in user settings (`~/.claude/settings.json`, not project settings), with `--settings <file>` or in managed settings: |
| 24 | |
| 25 | ```json |
Skill EnumerationAgent Snooping · README.md:36
| 34 | ``` |
| 35 | |
| 36 | It is written to `.claude/skills/webfetch-cache/`, which Claude Code auto-loads as `webfetch-cache@skills-dir`. For one session with hot reload: `claude --plugin-dir .claude/skills/webfetch-cache`. `claude plugin validate .claude/skills/web |
| 37 | |
| 38 | **Requirements.** Mods are on by default in Claude Code 2.1.287+. Typed against Anthropic's declarations: https://github.com/anthropics/claude-code/tree/main/mods |
npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.MCP Rug Pull · README.md:32
| 30 | |
| 31 | ```sh |
| 32 | npx claude-code-templates@latest --mod productivity/webfetch-cache |
| 33 | claude |
| 34 | ``` |
Without declared permissions the skill's intent is opaque and cannot be validated.MCP Least Privilege · SKILL.md:1
No code excerpt: this observation is about the component as a whole, not one line.
Learn more about NVIDIA SkillSpector or how components are reviewed.