rockyCAUTION
A blind alien engineer who listens to your terminal, never forgets, and makes sure you still understand your own code.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://github.com/poggufanz/rocky/actions/workflows/ci.yml) [](https://www.npmjs.com/package/@poggufanz/rocky-cli) [](LICENSE)
Release v1.0.1 | Changelog | License | Security | Contributing
Your AI writes code. Rocky makes sure you still understand what you own. He is a blind engineer who lives in your terminal, and he remembers failed commands and what fixed them, then brings that history back when the same trouble returns — to you, and through MCP to the AI agent working beside you, so it stops retrying what already failed on this machine. Supported agent hooks can also keep a bounded record of what you asked for, which files changed, and why the agent said it changed them. He is a witness, not a judge: he never diagnoses a failure or proposes a fix he does not remember, and docs/character.md holds the rules that boundary comes from.
The useful part is local and deterministic. Rocky has no daemon, no telemetry, and zero runtime dependencies. Optional AI uses an Ollama instance that you run on loopback.
You teach, I remember. I remind, you understand. This is good trade.
See it work
0d87b1d57253OBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add rocky-cli --env OPENROUTER_API_KEY=${OPENROUTER_API_KEY} --env PASSWORD=${PASSWORD} --env TYPESAFE_API_KEY=${TYPESAFE_API_KEY} -- npx -y @poggufanz/[email protected]{
"mcpServers": {
"rocky-cli": {
"command": "npx",
"args": [
"-y",
"@poggufanz/[email protected]"
],
"env": {
"OPENROUTER_API_KEY": "${OPENROUTER_API_KEY}",
"PASSWORD": "${PASSWORD}",
"TYPESAFE_API_KEY": "${TYPESAFE_API_KEY}"
}
}
}
}Exposed tools (14)
12 read · 1 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
activity_for_file | read | Evidence chain for one relative path in a consented repo: metadata only, weak links stay candidates. Example: {\ |
activity_recent | read | Newest listening events with metadata only: locators, hashes, evidence IDs, edge basis, coverage. Example: {\ |
bundle_get | read | Graph subview for one bundle: per-file evidence, sources, edge bases, coverage. Example: {\ |
bundles_list | write | Work-episode and commit bundles from exact IDs only, never time similarity. Example: {\ |
fetch_record | read | Fetch full detail for one id returned by search_knowledge. Example: {\ |
recall | read | Search remembered failures and fixes. |
recall_with_ai | read | Search remembered failures with optional AI ranking. |
recent_failures | read | List recent remembered failures. |
rocky-voice | read | Use when the user asks Rocky to narrate or interpret Rocky recall, failure, fix, or stats output. |
search_knowledge | read | Search remembered failures, fixes, AND agent-change knowledge (user intent, agent rationale, changed files). |
session_timeline | read | Graph events for one native session with pagination, metadata only. Example: {\ |
stats | destructive | Read bounded memory statistics and coverage metadata: scanned, skipped, truncated, and memory version. |
teach_lookup | read | Read-only teach card for a path and optional line or snippet: a remembered witness explanation when one matches, else an assembled evidence ladder. Example: {\ |
why_file | read | Recent remembered reasons agents changed one file, newest first, with separate incomplete-coverage disclosure. Example: {\ |
Trust audit
CAUTIONgrade F · trust 55/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (6 observation(s))
- Network
- declared (10 observation(s))
- Shell
- declared (5 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- found
Findings (25)
export const OLLAMA_ORIGIN = "http://127.0.0.1:11434";
return `http://127.0.0.1:${port}`;? `http://127.0.0.1:${handle.port}/?v=dash#${handle.token}`? `http://127.0.0.1:${handle.port}/?v=listening#${handle.token}`const url = new URL(request.url ?? "/", `http://127.0.0.1:${port}`);text: "naikin dikit sk-ant-abcdefghijklmnopqrst123\n\u001b[31m\u202e",
{ v: 1, agent: "codex", kind: "mechanism", ts: 1, tool: "Edit", path: "src/sk-ant-abcdefghijklmnopqrst123\u001b[31m\u202e.ts", excerpt: "color: red" },{ v: 1, agent: "codex", kind: "mechanism", ts: 2, tool: "Edit", path: "src/sk-ant-abcdefghijklmnopqrst456.ts", excerpt: "background: blue" },"src/sk-ant-abcdefghijklmnopqrst123.ts",
"src/sk-ant-abcdefghijklmnopqrst456.ts",
assert.ok(!JSON.stringify(triple).includes("AKIAABCDEFGHIJKLMNOP"));assert.equal(log.includes("AKIAABCDEFGHIJKLMNOP"), false);const secret = "AKIAABCDEFGHIJKLMNOP";
"AKIAABCDEFGHIJKLMNOP",
const secret = "AKIAABCDEFGHIJKLMNOP";
tags: [modernKey, 'api_key="api-aB3dE5fG7hI9jK2mN4pQ6"'],
const token = "sk-ant-abcdefghijklmnopqrst123";
const token = "sk-ant-abcdefghijklmnopqrst123";
const token = "sk-ant-abcdefghijklmnopqrst123";
const token = "sk-ant-abcdefghijklmnopqrst123";
"ghp_abcdefghijklmnopqrstuvwxyz1234567890",
"github_pat_abcdefghijklmnopqrstuvwxyz123456",
writeFileSync(join(dir, "secret.txt"), "const token = \"ghp_123456789012345678901234567890123456\";\n");
assert.ok(!got.diff.includes("ghp_123456789012345678901234567890123456"));assert.ok(!truncated.diff.includes("ghp_123456789012345678901234567890123456"));Gates applied: no_behavioural_pass.
0d87b1d57253full audit observations/trust-audit/mcp-server/poggufanz__rocky.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 0d87b1d57253 | CAUTION | F | 55 | first audit |
Questions
What is the rocky MCP server?
A blind alien engineer who listens to your terminal, never forgets, and makes sure you still understand your own code.
What tools does rocky expose?
14 in total: 12 read-only, 1 that write, and 1 that can delete or overwrite (stats). Every one is listed on this page with its risk.
Is rocky safe to connect to an agent?
With care. The audit graded it F (55/100) and found 25 things worth knowing before you trust this server, listed below with the exact line each was found on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does rocky need?
It reads OPENROUTER_API_KEY, PASSWORD and TYPESAFE_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How current is this page?
The grade is for one exact copy of the source (0d87b1d57253), read on 2026-10-07. The repository is watched and re-audited when it changes.