fix: wire up --prompt-caching flag, curate flags, and fix doc/env-var inaccuracies

- Add --prompt-caching CLI flag for odek run/repl/serve
- Add ODEK_PROMPT_CACHING env var and prompt_caching config key
- Wire up --apply and --interactive flags for 'odek skill curate'
- Fix SANDBOXING.md: KODE_SANDBOX_* -> ODEK_SANDBOX_* (wrong prefix)
- Fix odek-session-management skill: session cleanup uses UpdatedAt, not file mtime
- Fix odek-prompt-caching skill: config key is prompt_caching, not cache
- Remove non-existent --sse-addr from MCP docs/skills
- Add ODEK_PROMPT_CACHING to env var tables in CONFIG.md and CHEATSHEET.md

Author: molty3000

cmd/odek/main.go

@@ -235,6 +235,7 @@ type runFlags struct {
 	Sandbox  *bool // nil = not set
 	NoColor  *bool // nil = not set
 	NoAgents *bool // nil = not set
+	PromptCaching *bool // nil = not set; true = enable prompt caching
 	Session  *bool // nil = not set; true = save session after run
 	Learn    *bool // nil = not set; true = enable skills learning mode
 	Task     string
@@ -300,6 +301,9 @@ func parseRunFlags(args []string) (runFlags, error) {
 		case "--no-agents":
 			f.NoAgents = boolPtr(true)
 			i++
+		case "--prompt-caching":
+			f.PromptCaching = boolPtr(true)
+			i++
 		case "--session":
 			f.Session = boolPtr(true)
 			i++
@@ -353,6 +357,7 @@ type replFlags struct {
 	Model    string
 	Thinking string
 	Sandbox  *bool // nil = not set
+	PromptCaching *bool // nil = not set; true = enable prompt caching
 
 	// Sandbox-specific CLI flags
 	SandboxImage    string
@@ -415,6 +420,9 @@ func parseReplFlags(args []string) (replFlags, error) {
 		case "--sandbox-user":
 			f.SandboxUser = args[i+1]
 			i += 2
+		case "--prompt-caching":
+			f.PromptCaching = boolPtr(true)
+			i++
 		default:
 			// Unrecognized flag or positional — skip it
 			i++
@@ -443,7 +451,7 @@ Commands:
   run --session       Execute and save conversation as a session
   continue            Continue the most recent session (or by --id)
   repl                Interactive REPL mode (multi-turn session)
-                       Accepts --model, --thinking, --sandbox, and
+                       Accepts --model, --thinking, --sandbox, --prompt-caching, and
                        --sandbox-* flags just like odek run.
   serve               Web UI server with WebSocket streaming
                        Open http://localhost:8080 in your browser.
@@ -476,6 +484,7 @@ Run flags:
   --temperature <n>    LLM temperature 0.0–2.0 (default: 0 = deterministic)
   --no-color           Disable colored terminal output
   --no-agents          Skip loading AGENTS.md from working directory
+  --prompt-caching     Enable prompt caching markers (Anthropic/DeepSeek/OpenAI)
   --session            Save conversation as a multi-turn session
   --learn              Enable skill learning mode — on by default, no flag needed
   --no-learn           Disable skill learning mode (overrides config/default)
@@ -488,6 +497,7 @@ Skill commands:
   odek skill import <uri> [flags]    Import a skill from file:// or https://
                                      Flags: --basic (skip LLM), --yes (auto-approve)
   odek skill curate                  Analyze skills for quality, staleness, overlap
+                                     Flags: --apply (apply changes), --interactive (review one-by-one)
 
 Sandbox flags:
   --sandbox            Run in isolated Docker container
@@ -782,6 +792,7 @@ func run(args []string) error {
 		Sandbox:  f.Sandbox,
 		NoColor:  f.NoColor,
 		NoAgents: f.NoAgents,
+		PromptCaching: f.PromptCaching,
 		Learn:    f.Learn,
 		System:   f.System,
 		Task:     f.Task,
@@ -906,6 +917,7 @@ func run(args []string) error {
 		Renderer:       rend,
 		Skills:         skillsCfg,
 		SkillManager:   sm,
+		PromptCaching:  resolved.PromptCaching,
 	})
 	if err != nil {
 		return err
@@ -1628,11 +1640,27 @@ func skillCmd(args []string) error {
 		return nil
 
 	case "curate":
+		// Parse --apply and --interactive flags
+		apply := false
+		interactive := false
+		var remainingArgs []string
+		for _, arg := range subArgs {
+			switch arg {
+			case "--apply":
+				apply = true
+			case "--interactive":
+				interactive = true
+			default:
+				remainingArgs = append(remainingArgs, arg)
+			}
+		}
+		_ = remainingArgs // future use: filter by skill name
 		sm := skills.NewSkillManager(userDir, "./.odek/skills")
 		allSkills := append(sm.Result.AutoLoad, sm.Result.Lazy...)
 		report := skills.CurateSkills(allSkills, skills.CurateOptions{
 			StalenessDays: 90,
-			Apply:         false,
+			Apply:         apply,
+			Interactive:   interactive,
 		})
 		fmt.Print(skills.FormatCurationReport(report))
 		return nil
@@ -1801,6 +1829,7 @@ func continueCmd(args []string) error {
 		Renderer:       rend,
 		Skills:         skillsCfg,
 		SkillManager:   sm,
+		PromptCaching:  resolved.PromptCaching,
 	})
 	if err != nil {
 		return err

cmd/odek/repl.go

@@ -50,6 +50,7 @@ func replCmd(args []string) error {
 		Model:    f.Model,
 		Thinking: f.Thinking,
 		Sandbox:  f.Sandbox,
+		PromptCaching: f.PromptCaching,
 
 		SandboxImage:    f.SandboxImage,
 		SandboxNetwork:  f.SandboxNetwork,
@@ -145,6 +146,7 @@ func replCmd(args []string) error {
 		Skills:         skillsCfg,
 		SkillManager:   sm,
 		MemoryConfig:   resolved.Memory,
+		PromptCaching:  resolved.PromptCaching,
 	})
 	if err != nil {
 		return err

cmd/odek/serve.go

@@ -35,6 +35,7 @@ func serveCmd(args []string) error {
 	// Sandbox CLI flags (nil pointers = not set)
 	var sandbox *bool
 	var sandboxReadonly *bool
+	var promptCaching *bool
 	var sandboxImage, sandboxNetwork, sandboxMemory, sandboxCPUs, sandboxUser string
 
 	for i := 0; i < len(args); i++ {
@@ -78,13 +79,16 @@ func serveCmd(args []string) error {
 			if i < len(args) {
 				sandboxUser = args[i]
 			}
+		case "--prompt-caching":
+			promptCaching = boolPtr(true)
 		default:
 			return fmt.Errorf("unknown flag %q for serve", args[i])
 		}
 	}
 
 	resolved := config.LoadConfig(config.CLIFlags{
 		Sandbox:         sandbox,
+		PromptCaching:   promptCaching,
 		SandboxImage:    sandboxImage,
 		SandboxNetwork:  sandboxNetwork,
 		SandboxReadonly: sandboxReadonly,

docs/CHEATSHEET.md

@@ -12,7 +12,6 @@ odek serve                           # Web UI (http://127.0.0.1:8080)
 odek serve --open                    # Web UI + auto-open browser
 odek subagent --goal "review auth"   # Spawn subagent
 odek mcp                             # Expose tools via MCP stdio
-odek mcp --sse-addr :8081            # Expose via SSE
 
 # Sandbox flags (apply to run/repl/serve)
 odek run --sandbox "build safely"
@@ -197,6 +196,7 @@ odek mcp --sse-addr :8081                   # SSE transport
 | `ODEK_SYSTEM` | system |
 | `ODEK_NO_COLOR` | no_color |
 | `ODEK_NO_AGENTS` | no_agents |
+| `ODEK_PROMPT_CACHING` | prompt_caching |
 | `ODEK_MAX_CONCURRENCY` | max_concurrency |
 | `ODEK_CTX` | ctx (comma-separated file paths) |
 | `ODEK_API_KEY` | api_key (preferred) |

docs/CLI.md

@@ -32,7 +32,7 @@
 
 | Flag | Type | Default | Description |
 |------|------|---------|-------------|
-| `--model <name>` | string | `deepseek-chat` | LLM model — profiles auto-set thinking/timeout (see [Providers](docs/PROVIDERS.md)) |
+| `--model <name>` | string | `deepseek-chat` | LLM model — profiles auto-set thinking/timeout (see [Providers](docs/PROVIDERS.md)). Consider using `deepseek-v4-flash` for faster/cheaper tasks. |
 | `--base-url <url>` | string | `https://api.deepseek.com/v1` | OpenAI-compatible API endpoint |
 | `--max-iter <n>` | int | `90` | Max think→act cycles |
 | `--thinking <level>` | string | profile default | Reasoning depth: `enabled`/`disabled`/`low`/`medium`/`high` |

docs/CONFIG.md

@@ -63,6 +63,7 @@ Every config knob has a `ODEK_*` counterpart:
 | `ODEK_NO_AGENTS` | `--no-agents` | bool |
 | `ODEK_SYSTEM` | `--system` | string |
 | `ODEK_SKILLS_LEARN` | `skills.learn` | bool |
+| `ODEK_PROMPT_CACHING` | `prompt_caching` | bool |
 | `ODEK_SANDBOX_IMAGE` | `--sandbox-image` | string |
 | `ODEK_SANDBOX_NETWORK` | `--sandbox-network` | string |
 | `ODEK_SANDBOX_READONLY` | `--sandbox-readonly` | bool |

docs/SANDBOXING.md

@@ -20,7 +20,7 @@ odek run --sandbox "protoc --version"
 
 ## Config reference
 
-All sandbox settings are available in `~/.odek/config.json`, `./kode.json`, `KODE_*` env vars, and CLI flags, following the same [priority chain](CONFIG.md).
+All sandbox settings are available in `~/.odek/config.json`, `./kode.json`, `ODEK_*` env vars, and CLI flags, following the same [priority chain](CONFIG.md).
 
 ### Config file fields
 
@@ -47,28 +47,28 @@ All sandbox settings are available in `~/.odek/config.json`, `./kode.json`, `KOD
 
 | Field | Env var | CLI flag | Type | Default | Description |
 |-------|---------|----------|------|---------|-------------|
-| `sandbox` | `KODE_SANDBOX` | `--sandbox` | bool | `false` | Enable/disable sandbox isolation |
-| `sandbox_image` | `KODE_SANDBOX_IMAGE` | `--sandbox-image` | string | `alpine:latest` | Docker image for the sandbox container |
-| `sandbox_network` | `KODE_SANDBOX_NETWORK` | `--sandbox-network` | string | `bridge` | Docker network mode |
-| `sandbox_readonly` | `KODE_SANDBOX_READONLY` | `--sandbox-readonly` | bool | `false` | Mount working directory read-only |
-| `sandbox_memory` | `KODE_SANDBOX_MEMORY` | `--sandbox-memory` | string | `""` | Memory limit (e.g. `512m`, `2g`) |
-| `sandbox_cpus` | `KODE_SANDBOX_CPUS` | `--sandbox-cpus` | string | `""` | CPU limit (e.g. `0.5`, `2`) |
-| `sandbox_user` | `KODE_SANDBOX_USER` | `--sandbox-user` | string | `""` | Run as user (`uid:gid` or name) |
+| `sandbox` | `ODEK_SANDBOX` | `--sandbox` | bool | `false` | Enable/disable sandbox isolation |
+| `sandbox_image` | `ODEK_SANDBOX_IMAGE` | `--sandbox-image` | string | `alpine:latest` | Docker image for the sandbox container |
+| `sandbox_network` | `ODEK_SANDBOX_NETWORK` | `--sandbox-network` | string | `bridge` | Docker network mode |
+| `sandbox_readonly` | `ODEK_SANDBOX_READONLY` | `--sandbox-readonly` | bool | `false` | Mount working directory read-only |
+| `sandbox_memory` | `ODEK_SANDBOX_MEMORY` | `--sandbox-memory` | string | `""` | Memory limit (e.g. `512m`, `2g`) |
+| `sandbox_cpus` | `ODEK_SANDBOX_CPUS` | `--sandbox-cpus` | string | `""` | CPU limit (e.g. `0.5`, `2`) |
+| `sandbox_user` | `ODEK_SANDBOX_USER` | `--sandbox-user` | string | `""` | Run as user (`uid:gid` or name) |
 | `sandbox_env` | — | — | object | `{}` | Extra env vars injected into container |
 | `sandbox_volumes` | — | — | array | `[]` | Extra volume mounts (`host:container`) |
 
-> **Note:** `sandbox_env` and `sandbox_volumes` are config-file-only — they're too complex for flat env vars or CLI flags. For all other fields, env vars and CLI flags follow the standard `KODE_*` pattern.
+> **Note:** `sandbox_env` and `sandbox_volumes` are config-file-only — they're too complex for flat env vars or CLI flags. For all other fields, env vars and CLI flags follow the standard `ODEK_*` pattern.
 
 ### Env var examples
 
 ```bash
-KODE_SANDBOX=true \
-KODE_SANDBOX_IMAGE=python:3.12-slim \
-KODE_SANDBOX_NETWORK=none \
-KODE_SANDBOX_READONLY=true \
-KODE_SANDBOX_MEMORY=1g \
-KODE_SANDBOX_CPUS=4 \
-KODE_SANDBOX_USER=1000:1000 \
+ODEK_SANDBOX=true \
+ODEK_SANDBOX_IMAGE=python:3.12-slim \
+ODEK_SANDBOX_NETWORK=none \
+ODEK_SANDBOX_READONLY=true \
+ODEK_SANDBOX_MEMORY=1g \
+ODEK_SANDBOX_CPUS=4 \
+ODEK_SANDBOX_USER=1000:1000 \
   odek run "process untrusted data"
 ```
 

internal/config/loader.go

@@ -49,6 +49,10 @@ type CLIFlags struct {
 	Learn    *bool // nil = not set
 	Task     string
 
+	// PromptCaching enables prompt caching markers for supported providers.
+	// Config: prompt_caching, ODEK_PROMPT_CACHING, --prompt-caching.
+	PromptCaching *bool // nil = not set
+
 	// Sandbox-specific
 	SandboxImage    string
 	SandboxNetwork  string
@@ -92,6 +96,9 @@ type FileConfig struct {
 	NoColor  *bool `json:"no_color,omitempty"`
 	NoAgents *bool `json:"no_agents,omitempty"`
 
+	// PromptCaching enables prompt caching markers for supported providers.
+	PromptCaching *bool `json:"prompt_caching,omitempty"`
+
 	System string `json:"system,omitempty"`
 
 	// Sandbox-specific fields.
@@ -158,6 +165,7 @@ type ResolvedConfig struct {
 	Sandbox      bool
 	NoColor      bool
 	NoAgents     bool
+	PromptCaching bool
 	System       string
 
 	// SandboxImage is the Docker image for the sandbox container.
@@ -380,6 +388,9 @@ func LoadConfig(cli CLIFlags) ResolvedConfig {
 	if v := envBool("NO_AGENTS"); v != nil {
 		cfg.NoAgents = v
 	}
+	if v := envBool("PROMPT_CACHING"); v != nil {
+		cfg.PromptCaching = v
+	}
 	if v := envString("SYSTEM"); v != "" {
 		cfg.System = v
 	}
@@ -456,6 +467,9 @@ func LoadConfig(cli CLIFlags) ResolvedConfig {
 	if cli.NoAgents != nil {
 		cfg.NoAgents = cli.NoAgents
 	}
+	if cli.PromptCaching != nil {
+		cfg.PromptCaching = cli.PromptCaching
+	}
 	if cli.Learn != nil {
 		if cfg.Skills == nil {
 			cfg.Skills = &SkillsConfig{}
@@ -532,6 +546,9 @@ func LoadConfig(cli CLIFlags) ResolvedConfig {
 	if cfg.NoAgents != nil {
 		resolved.NoAgents = *cfg.NoAgents
 	}
+	if cfg.PromptCaching != nil {
+		resolved.PromptCaching = *cfg.PromptCaching
+	}
 	if cfg.SandboxReadonly != nil {
 		resolved.SandboxReadonly = *cfg.SandboxReadonly
 	}