chore: fix documentation drift, inaccurate comments, and unit test noise

cli/README.md

@@ -303,6 +303,7 @@ The main command groups are:
 - `osb command`: command execution and persistent sessions
 - `osb file`: file and directory operations
 - `osb egress`: runtime egress policy
+- `osb credential-vault`: manage the Credential Vault
 - `osb diagnostics`: stable diagnostics logs and events
 - `osb devops`: experimental legacy diagnostics
 - `osb config`: local CLI configuration
@@ -327,6 +328,7 @@ Bundled skills:
 - `command-execution`
 - `file-operations`
 - `network-egress`
+- `credential-vault`
 - `sandbox-troubleshooting`
 
 Supported targets:

cli/src/opensandbox_cli/commands/command.py

@@ -131,7 +131,13 @@ def _handle_execution_error(obj: ClientContext, execution) -> None:
 @click.option("-d", "--background", is_flag=True, default=False, help="Run in background.")
 @click.option("-w", "--workdir", default=None, help="Working directory.")
 @click.option("-t", "--timeout", type=DURATION, default=None, help="Command timeout (e.g. 30s, 5m).")
-@output_option("table", "json", "yaml", "raw")
+@output_option(
+    "table",
+    "json",
+    "yaml",
+    "raw",
+    help_text="Output format. Foreground mode: raw only (default). Background mode: table, json, or yaml.",
+)
 @click.pass_obj
 @handle_errors
 def command_run(

cli/src/opensandbox_cli/commands/sandbox.py

@@ -51,9 +51,6 @@ def sandbox_group(ctx: click.Context) -> None:
         click.echo(ctx.get_help())
 
 
-# Alias: osb sb ...
-sandbox_group.name = "sandbox"
-
 _SANDBOX_STATE_CANONICAL = {
     state.lower(): state for state in SandboxState.values()
 }

cli/src/opensandbox_cli/commands/skills.py

@@ -184,6 +184,7 @@ class TargetConfig(TypedDict):
     "command-execution": "Execution",
     "file-operations": "Files",
     "network-egress": "Network",
+    "credential-vault": "Credentials",
     "sandbox-troubleshooting": "Troubleshooting",
 }
 
@@ -418,9 +419,6 @@ def _install_guidance_text() -> str:
         "    osb skills install <skill-name> --target <tool> --scope <scope>\n\n"
         "  Install all bundled skills for one tool:\n"
         "    osb skills install --all-builtins --target <tool> --scope <scope>\n\n"
-        "  Discover skills and targets:\n"
-        "    osb skills list\n"
-        "    osb skills show <skill-name>\n\n"
         f"  Available skills: {', '.join(_ALL_SKILL_NAMES)}\n"
         f"  Available targets: {', '.join(_ALL_TARGET_NAMES)}\n"
         f"  Available scopes: {', '.join(_ALL_SCOPE_NAMES)}"

cli/src/opensandbox_cli/utils.py

@@ -148,7 +148,7 @@ def select_output_format(
     allowed: tuple[str, ...],
     fallback: str,
 ) -> str:
-    """Resolve a command-scoped output format from explicit input, config, and fallback."""
+    """Resolve a command-scoped output format from explicit input and fallback."""
     if requested:
         if requested not in allowed:
             allowed_list = ", ".join(allowed)

cli/tests/test_commands.py

@@ -72,7 +72,7 @@ def _make_output(fmt: str) -> OutputFormatter:
     ctx.make_output.side_effect = _make_output
     ctx.get_manager.return_value = manager or MagicMock()
     ctx.connect_sandbox.return_value = sandbox or MagicMock()
-    ctx.resolve_sandbox_id.side_effect = lambda prefix: prefix  # passthrough
+    ctx.resolve_sandbox_id.side_effect = lambda prefix: prefix
     ctx.connection_config = MagicMock()
     ctx.close = MagicMock()
     return ctx

components/egress/Dockerfile

@@ -124,10 +124,13 @@ RUN chown mitmproxy:mitmproxy /var/lib/mitmproxy/.mitmproxy/config.yaml \
 COPY --from=builder /out/egress /opt/opensandbox-egress/egress
 COPY --from=builder /out/opensandbox-supervisor /opt/opensandbox-egress/supervisor
 # Pre-start hook: reap any mitmdump left over from a previous crashed
-# egress so the new launch can bind the transparent-MITM listen port.
-# Intentionally does NOT touch iptables/nft rules — the sidecar shares
-# a network namespace with the workload, so leaving rules in place keeps
-# egress filtering active across the supervisor's backoff window.
+# egress so the new launch can bind the transparent-MITM listen port, and
+# remove stale redirect rules (iptables DNS 53 redirect, transparent-MITM
+# REDIRECT, and the opensandbox_dns_redirect nft table) left pointing at the
+# dead processes. It intentionally leaves the `inet opensandbox` policy
+# table untouched — the sidecar shares a network namespace with the
+# workload, so leaving the filter rules in place keeps egress enforcement
+# active across the supervisor's backoff window.
 COPY components/egress/scripts/cleanup.sh /opt/opensandbox-egress/cleanup.sh
 RUN chmod 0755 /opt/opensandbox-egress/cleanup.sh \
             /opt/opensandbox-egress/egress \

components/egress/docs/opentelemetry.md

@@ -29,7 +29,7 @@ Do not drop them: the instrument records **seconds**, while the SDK default boun
 the spec's millisecond ladder (`0, 5, 10, … 10000`), so every realistic latency would fall
 into the single `le=5` bucket and the quantiles would be meaningless.
 
-The head resolves a cache hit (sub-millisecond) up to one upstream timeout
+The head spans a fast single-upstream exchange (sub-millisecond) up to one upstream timeout
 (`OPENSANDBOX_EGRESS_DNS_UPSTREAM_TIMEOUT`, 5s by default). The coarse tail exists because
 the recorded duration covers the **whole resolver chain**: forwarding walks the upstreams
 serially, each with the full timeout, so a query can legitimately take
@@ -84,12 +84,11 @@ All egress metrics may include shared attributes:
 
 ## OTEL Endpoint Configuration
 
-Metric export is enabled only when at least one OTLP endpoint is set.
+Metric export is enabled when an OTLP endpoint is set, or when a node IP is available via `HOST_IP` or `/etc/hostinfo` (metrics are then exported insecurely to `<ip>:4318`).
 
 - `OTEL_EXPORTER_OTLP_METRICS_ENDPOINT` (preferred)
 - `OTEL_EXPORTER_OTLP_ENDPOINT` (fallback)
-
-If both are unset, egress keeps metrics local (no OTLP export).
+- `HOST_IP` (fallback endpoint host when both OTLP env vars are unset)
 
 ### Minimal Example
 

components/egress/nameserver_test.go

@@ -37,7 +37,6 @@ func TestAllowIPsForNft_EmptyResolv(t *testing.T) {
 func TestAllowIPsForNft_ValidNameservers(t *testing.T) {
 	dir := t.TempDir()
 	resolv := filepath.Join(dir, "resolv.conf")
-	// Standard resolv.conf with two nameservers
 	content := "nameserver 192.168.65.7\nnameserver 10.0.0.1\n"
 	require.NoError(t, os.WriteFile(resolv, []byte(content), 0644))
 	ips := AllowIPsForNft(resolv)

components/egress/pkg/policy/policy_test.go

@@ -102,7 +102,6 @@ func TestWithExtraAllowIPs(t *testing.T) {
 	require.Len(t, allowV6, 1, "allowV6 length mismatch")
 	require.Equal(t, "2001:db8::1", allowV6[0])
 
-	// nil/empty ips returns same policy
 	require.Same(t, p, p.WithExtraAllowIPs(nil), "WithExtraAllowIPs(nil) should return same policy")
 	require.Same(t, p, p.WithExtraAllowIPs([]netip.Addr{}), "WithExtraAllowIPs([]) should return same policy")
 }

components/egress/pkg/telemetry/metrics.go

@@ -108,7 +108,7 @@ func registerEgressMetrics() error {
 		// boundaries are the spec's millisecond ladder (0, 5, 10, ... 10000), so every
 		// realistic DNS latency lands in the same bucket and the quantiles are noise.
 		//
-		// The head spans a cache hit (sub-ms) to one upstream timeout
+		// The head spans a fast single-upstream exchange (sub-ms) to one upstream timeout
 		// (DefaultDNSUpstreamTimeoutSec = 5s). The coarse tail covers the retry chain:
 		// forward() walks the resolvers serially, each with the full timeout, and the
 		// recorded duration is the whole chain — so a query can legitimately take

components/egress/policy_server.go

@@ -54,7 +54,8 @@ type nftApplier interface {
 	RemoveEnforcement(context.Context) error
 }
 
-// startPolicyServer: runtime POST/GET /policy, GET /healthz. nameserverIPs are merged into every nft
+// startPolicyServer serves the runtime policy API (GET/POST/PUT/PATCH/DELETE /policy,
+// /credential-vault* subroutes) and GET /healthz. nameserverIPs are merged into every nft
 // static apply so the pod’s resolv / private DNS still works alongside user egress rules.
 func startPolicyServer(
 	proxy policyUpdater,

components/egress/policy_utils.go

@@ -111,7 +111,8 @@ func removeRulesByTarget(rules []policy.EgressRule, targets []string) (kept, rem
 	return kept, removed
 }
 
-// mergeKey: domain targets lowercased for dedupe; IP/CIDR left as-is.
+// mergeKey lowercases targets for dedupe (IP/CIDR strings are lowered too,
+// which is lossless for canonical IP/CIDR forms).
 func mergeKey(r policy.EgressRule) string {
 	if r.Target == "" {
 		return r.Target

components/execd/RELEASE_NOTES.md

@@ -109,8 +109,8 @@ Thanks to these contributors ❤️
 - @csdbianhua
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.8
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.8
 
 # components/execd 1.0.7
 
@@ -144,8 +144,8 @@ Thanks to these contributors ❤️
 - @dependabot
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.7
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.7
 
 # components/execd 1.0.6
 
@@ -167,8 +167,8 @@ Thanks to these contributors ❤️
 - @dependabot
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.6
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.6
 
 # components/execd 1.0.5
 
@@ -185,8 +185,8 @@ Thanks to these contributors ❤️
 - @Pangjiping
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.5
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.5
 
 # components/execd 1.0.4
 
@@ -209,8 +209,8 @@ Thanks to these contributors ❤️
 - @ninan-nn
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.4
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.4
 
 # components/execd 1.0.3
 
@@ -233,8 +233,8 @@ Thanks to these contributors ❤️
 - @jwx0925
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.3
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.3
 
 # components/execd 1.0.2
 
@@ -262,8 +262,8 @@ Thanks to these contributors ❤️
 - @ninan-nn
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.2
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.2
 
 # components/execd 1.0.1
 
@@ -295,8 +295,8 @@ Thanks to these contributors ❤️
 - @jwx0925
 
 ---
-- Docker Hub: opensandbox/execd:v1.0.9
-- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.9
+- Docker Hub: opensandbox/execd:v1.0.1
+- Aliyun Registry: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.1
 
 # components/execd 1.0.0