Files
ee684a29a9 feat(sandbox): presets — additive workload allowance bundles (#387)
* feat(sandbox): introduce presets - additive workload allowance bundles

Presets are named, additive-only bundles of sandbox allowances for a
specific workload (git hooks tooling, Astro/Vite/Next.js dev servers).
They solve the per-workload tuning friction from #384 without weakening
the default posture: no built-in profile references a preset, presets
cannot carry deny rules or profile booleans (strict YAML decoding), and
mandatory denies still win everywhere except the existing exact-match
suppression.

- Preset schema with metadata (author, labels) and schema_version gating
- Registry over ordered sources (embedded builtin, user dir); builtin
  wins name collisions; source abstraction is the extension point for a
  future hosted registry and SafeDep cloud sync
- Official presets: git, astro, vite, nextjs (with threat notes)
- Overlay/runtime integration: pmg sandbox allow preset=<name> and
  --sandbox-allow preset=<name>, stored by reference, resolved at apply
  time, missing presets warn (fail closed) instead of aborting
- Profile integration: presets: [...] list resolved after inherits
- CLI: pmg sandbox preset list (metadata filters, --json), show (prints
  YAML with threat notes), lint
- Docs: user guide (docs/sandbox-presets.md) and design spec

Closes #384

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* fix(sandbox): address review findings on presets

- Presets never modify deny lists: a profile authored deny now survives a
  preset allowing the same path (deny-beats-allow keeps it enforced).
  Regression test added.
- Profile inspection commands (show, diff, lint) construct the profile
  registry with the user-aware preset registry so they agree with runtime
  resolution of custom profiles referencing user presets.
- Handle stderr write error when warning about unresolvable presets.
- Compute preset show underline from the uncolored header.
- Use path.Join for embed.FS reads (slash-separated on all platforms).
- Clarify in docs that lint-staged/astro are examples of preset workloads.
- Drop the design spec from the PR per review.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* fix(sandbox): harden preset precedence against authored denies

Addresses external security review findings on the preset mechanism:

- Bubblewrap: a mandatory write-denied path listed in allow_read lost its
  protection when a later writable parent bind covered it (bwrap last
  mount wins) - exactly the git preset shape (allow_read .git/config +
  allow_write .git/**). The mandatory deny now re-binds the path
  read-only after all writable mounts instead of being skipped.
  Regression test asserts mount ordering. Landlock and Seatbelt were
  unaffected (tests added for the same policy shape on Landlock).
- Environment: ScrubEnv is allow-wins, so a preset environment allowance
  could override a profile-authored deny. Preset env allowances
  overlapping an authored deny pattern are now dropped at application
  time (conservative bidirectional glob overlap, fail closed). Surviving
  entries still opt out of built-in credential scrubbing as intended.
- Network: removed allow_outbound from the preset schema. Both platform
  translators are all-or-nothing for outbound (one allow rule means
  blanket network access), so a preset outbound entry would silently
  change network posture far beyond what its YAML conveys. Strict
  decoding rejects the key.
- Added a dual-path expansion equivalence test (profile presets: field
  vs overlay/--sandbox-allow) and documented the precedence guarantees
  in docs/sandbox-presets.md.

Explicit --sandbox-allow and pmg sandbox allow overrides keep their
existing semantics.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* docs(sandbox): document env and preset allowances in allow command and overlay docs

pmg sandbox allow help, the --sandbox-allow flag usage, and the project
overlay docs enumerated only read/write/exec/net types. Add env and
preset to all of them, with an overlay example for persisting an env
allowance and a note on why env entries are not auto-promoted by --last.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* chore(sandbox): trim preset code comments to corner cases and minimal godocs

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* fix(sandbox): exact glob intersection for preset env deny overlap

The bidirectional literal-text heuristic missed overlapping globs with
different literal structure: preset allow AWS_*_KEY and authored deny
AWS_SECRET_* both match AWS_SECRET_ACCESS_KEY but neither pattern
matches the other's text, so the allowance merged and allow-wins
scrubbing exposed the variable. EnvPatternsOverlap now computes exact
intersection non-emptiness for the name glob dialect (case-insensitive,
'*' any sequence, '?' single char) via memoized DP.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* fix(sandbox): preset env allowances are exact names, not globs

Glob-vs-glob intersection is a losing game: every dialect extension
(character classes today) silently reopens the deny-bypass hole.
Restricting preset environment allowances to literal variable names
makes the authored-deny precedence check exact by construction: each
deny pattern is evaluated against the concrete name with the same
matcher ScrubEnv uses at runtime, so the decision cannot diverge from
enforcement regardless of deny dialect. Removes the glob intersection
machinery. Profile and --sandbox-allow env globs are unaffected.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* fix(sandbox): reject mandatory-deny targets in preset paths

Preset validation relied on IsSensitiveProjectTarget, which covers fewer
files than util.DANGEROUS_FILES. A preset naming .git-credentials,
.pgpass, .docker/config.json or .config/gh exactly would exact-match
suppress the mandatory deny; .git/config in allow_write would suppress
the write protection. Preset paths are now checked against
DANGEROUS_FILES (single source of truth), .git/hooks is rejected in any
direction, and .git/config is rejected for write/exec while read stays
allowed for git repo discovery. Docs state the two deliberate opt-outs
precisely.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* feat(sandbox): preset init and edit commands for community authoring

pmg sandbox preset init scaffolds a valid user preset (metadata flags,
threat-note template, starter rule) and refuses built-in names since
builtins win resolution. pmg sandbox preset edit opens the file via the
shared editor package and validates the result, warning when a user
preset is shadowed by a built-in. Docs lead with the scaffolded flow and
spell out builtin-vs-community provenance in preset list.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

* refactor(sandbox): move mandatory-target matching into util

Preset path validation re-encoded knowledge util already owns: the
dangerous-files comparison and hardcoded .git/config and .git/hooks
strings. util now exports GitConfigPath, GitHooksPath (also used by
GetMandatoryDenyPatterns), PathCoveredBy and DangerousFileMatch, and
preset validation consumes them so the mandatory deny policy has a
single definition.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PETCfE4crLcmodosz12qRn

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-07-21 15:14:07 +05:30

320 lines
10 KiB
Go

package sandbox
import (
"context"
"fmt"
"net"
"os/exec"
"strconv"
"github.com/safedep/dry/usefulerror"
"github.com/safedep/dry/utils"
"github.com/safedep/pmg/errcodes"
)
// ExecutionContext carries runtime data known only at spawn time.
type ExecutionContext struct {
// ProxyAddr is the loopback TCP address of the running PMG proxy
// (e.g. "127.0.0.1:54321"). Empty when no proxy flow is active.
ProxyAddr string
}
// ValidateNetworkLockdown enforces the network_via_proxy_only fail-closed contract
// for drivers that support it. Returns the validated proxy port, or "" when
// lockdown is off.
func ValidateNetworkLockdown(policy *SandboxPolicy, rt *ExecutionContext) (string, error) {
if !utils.SafelyGetValue(policy.NetworkViaProxyOnly) {
return "", nil
}
if rt == nil || rt.ProxyAddr == "" {
return "", usefulerror.NewUsefulError().
WithCode(errcodes.SandboxRequiresProxy).
WithHumanError("network_via_proxy_only requires the PMG proxy flow").
WithHelp("This sandbox profile confines all network access to the PMG proxy, but no proxy is running.").
Wrap(fmt.Errorf("policy %s requires the PMG proxy flow: network_via_proxy_only is set but no proxy address was provided", policy.Name))
}
host, port, err := net.SplitHostPort(rt.ProxyAddr)
if err != nil {
return "", fmt.Errorf("network_via_proxy_only: proxy address %q is not host:port and not loopback: %w", rt.ProxyAddr, err)
}
ip := net.ParseIP(host)
if ip == nil || !ip.IsLoopback() {
return "", fmt.Errorf("network_via_proxy_only: refusing non-loopback proxy address %q", rt.ProxyAddr)
}
portNum, err := strconv.Atoi(port)
if err != nil || portNum < 1 || portNum > 65535 {
return "", fmt.Errorf("network_via_proxy_only: refusing non-numeric or out-of-range proxy port in %q", rt.ProxyAddr)
}
return port, nil
}
// DriverName identifies a sandbox driver implementation. Returned by
// Sandbox.Name() and used wherever code needs to refer to a specific driver.
type DriverName string
const (
DriverSeatbelt DriverName = "seatbelt"
DriverBubblewrap DriverName = "bubblewrap"
DriverLandlock DriverName = "landlock"
)
// ViolationKind is PMG's normalized taxonomy for sandbox denials.
type ViolationKind string
const (
ViolationKindFSRead ViolationKind = "fs_read"
ViolationKindFSWrite ViolationKind = "fs_write"
ViolationKindFSDeleteOrRename ViolationKind = "fs_delete_or_rename"
ViolationKindExec ViolationKind = "exec"
ViolationKindNetworkConnect ViolationKind = "network_connect"
ViolationKindNetworkBind ViolationKind = "network_bind"
ViolationKindGenericDeny ViolationKind = "generic_deny"
)
// ViolationReport is a best-effort sandbox violation summary collected from a
// sandbox implementation after command execution fails.
type ViolationReport struct {
SandboxName DriverName
PolicyName string
CorrelationID string
Violations []Violation
}
// Violation captures one sandbox denial event.
type Violation struct {
Kind ViolationKind
RawKind string
Target string
RuleTarget string
Process string
RawLog string
RuleLabel string
}
type violationReporter interface {
BestEffortViolation(err error) (*ViolationReport, error)
}
// ExecutionResult represents the result of executing a command in a sandbox.
// It contains sandbox internal state and allows for future extension with
// additional metadata (e.g., exit codes, resource usage, violation events).
// Callers must call Close() after cmd.Run() completes to clean up resources.
type ExecutionResult struct {
executed bool
sandbox Sandbox
scrubbedEnvCount int
}
// ExecutionResultOpt is a function that can be used to configure an ExecutionResult.
type ExecutionResultOpt func(*ExecutionResult)
// WithSandbox sets the sandbox for the ExecutionResult.
func WithExecutionResultSandbox(sb Sandbox) ExecutionResultOpt {
return func(r *ExecutionResult) {
r.sandbox = sb
}
}
// WithExecuted sets the executed flag for the ExecutionResult.
func WithExecutionResultExecuted(executed bool) ExecutionResultOpt {
return func(r *ExecutionResult) {
r.executed = executed
}
}
// NewExecutionResult creates a new ExecutionResult.
func NewExecutionResult(opts ...ExecutionResultOpt) *ExecutionResult {
r := &ExecutionResult{}
for _, opt := range opts {
opt(r)
}
return r
}
// ShouldRun returns true if the caller should execute cmd.Run().
func (r *ExecutionResult) ShouldRun() bool {
return !r.executed
}
// SetScrubbedEnvCount records how many environment variables were scrubbed
// from the child process per the resolved environment policy.
func (r *ExecutionResult) SetScrubbedEnvCount(count int) {
r.scrubbedEnvCount = count
}
// ScrubbedEnvCount returns how many environment variables were scrubbed from
// the child process. Used to hint at scrubbing as a possible cause when the
// child fails.
func (r *ExecutionResult) ScrubbedEnvCount() int {
if r == nil {
return 0
}
return r.scrubbedEnvCount
}
// BestEffortViolation returns sandbox-specific best-effort violation details.
// Implementations may use platform logs or other weak signals, so callers
// should treat the result as advisory.
func (r *ExecutionResult) BestEffortViolation(err error) (*ViolationReport, error) {
if r == nil || r.sandbox == nil {
return nil, nil
}
reporter, ok := r.sandbox.(violationReporter)
if !ok {
return nil, nil
}
return reporter.BestEffortViolation(err)
}
// Close cleans up any resources allocated by the sandbox.
// Must be called after cmd.Run() completes.
func (r *ExecutionResult) Close() error {
if r.sandbox != nil {
return r.sandbox.Close()
}
return nil
}
// Sandbox represents a platform-specific sandbox executor that isolates
// package manager processes with controlled access to filesystem, network,
// and process execution resources.
type Sandbox interface {
// Execute prepares or runs a command in the sandbox with the given policy.
//
// Behavior varies by implementation:
// - CLI-based sandboxes (Seatbelt, Bubblewrap): Modify cmd in place by wrapping it
// with sandbox CLI (e.g., sandbox-exec). Returns ExecutionResult with executed=false.
// - Library-based sandboxes: Execute the command directly within the sandbox.
// Returns ExecutionResult with executed=true.
//
// Returns:
// - ExecutionResult: Contains execution state and metadata
// - error: Non-nil if sandbox setup or execution failed
//
// rt carries runtime data known only at spawn time; drivers treat a nil
// rt as &ExecutionContext{}.
//
// Callers must check result.ShouldRun() and only call cmd.Run() if true.
Execute(ctx context.Context, cmd *exec.Cmd, policy *SandboxPolicy, rt *ExecutionContext) (*ExecutionResult, error)
// Name returns the sandbox driver identifier.
Name() DriverName
// IsAvailable returns true if the sandbox is available and functional on this platform.
IsAvailable() bool
// Close cleans up any resources allocated by the sandbox (e.g., temporary files).
// Must be called after cmd.Run() completes. Idempotent - safe to call multiple times.
Close() error
}
// ProfileInfo describes a user profile discovered on disk.
type ProfileInfo struct {
// Name is the profile name, derived from the file name without extension.
Name string
// Path is the absolute path to the profile file.
Path string
// Shadowed is true when a built-in profile of the same name exists and
// will win during name-based resolution.
Shadowed bool
}
// ProfileSource identifies where a profile came from.
type ProfileSource string
const (
ProfileSourceBuiltin ProfileSource = "builtin"
ProfileSourceUser ProfileSource = "user"
)
// ProfileSummary describes a discoverable profile for listing purposes.
type ProfileSummary struct {
Name string
Source ProfileSource
Path string // "" for builtins; absolute path for user profiles
Inherits string
PackageManagers []string
Description string
Shadowed bool // true when a user file is masked by a same-named builtin
}
// ResolveOptions tunes variable expansion when resolving a policy for display
// or diffing. Zero values mean "use the current process environment".
type ResolveOptions struct {
CWD string
Home string
TmpDir string
}
// ProfileRegistry manages built-in and custom sandbox policies.
type ProfileRegistry interface {
// GetProfile retrieves a policy by name.
// Name can be a built-in profile (e.g., "npm-restrictive"), the bare name of a
// user profile under UserProfileDir(), or a path to a custom YAML file.
// Resolution order: built-ins first, then the user profile directory.
GetProfile(name string) (*SandboxPolicy, error)
// LoadCustomProfile loads a policy from a custom YAML file path.
LoadCustomProfile(path string) (*SandboxPolicy, error)
// ListProfiles returns all discoverable profiles: built-ins first, then
// user profiles (including shadowed entries).
ListProfiles() ([]ProfileSummary, error)
// ResolveProfile loads name and returns a policy with all path-bearing
// fields expanded against opts (or the process environment).
ResolveProfile(name string, opts ResolveOptions) (*SandboxPolicy, error)
// UserProfileDir returns the directory scanned for user profiles.
UserProfileDir() string
// ListUserProfiles enumerates *.yml / *.yaml files under UserProfileDir().
// A missing directory returns an empty slice with no error.
ListUserProfiles() ([]ProfileInfo, error)
// BuiltinProfileYAML returns the embedded YAML bytes for a built-in
// profile. Returns false if name is not a built-in.
BuiltinProfileYAML(name string) ([]byte, bool)
}
// RegistryOption configures a ProfileRegistry.
type RegistryOption func(*registryOptions)
type registryOptions struct {
userProfileDir string
presetRegistry PresetRegistry
}
// WithUserProfileDir sets the directory the registry uses to discover user
// profiles. The directory does not need to exist at construction time.
func WithUserProfileDir(dir string) RegistryOption {
return func(o *registryOptions) {
o.userProfileDir = dir
}
}
// WithPresetRegistry sets the preset registry used to expand `presets:`
// references in profiles. Defaults to a builtin-only preset registry.
func WithPresetRegistry(presets PresetRegistry) RegistryOption {
return func(o *registryOptions) {
o.presetRegistry = presets
}
}
// NewProfileRegistry creates a new profile registry with built-in policies.
func NewProfileRegistry(opts ...RegistryOption) (ProfileRegistry, error) {
return newDefaultProfileRegistry(opts...)
}