pwn 0.5.660 → 0.5.661

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: '06678f228c6acc69d9fb173f01f3825af6a7a8dcffcb8ed24c865bf281e0c65c'
-  data.tar.gz: 9967223f974d122e30cbb4be7aea369108e28b951ace091a1ed90d54825a4720
+  metadata.gz: a771af76d15bb7d7bf172b6c096f5902b5dd868793593e889b2376885cf84c19
+  data.tar.gz: 4beecdae03ebe816005ebdefa85d131bf12ea32c9e0585f759b25b11681412c7
 SHA512:
-  metadata.gz: 6feb604302f7e6fd428e8dcd2d1c6a888ed2e6d4c96f3e7b75b6537a5694c3053da6a69da751275c8c168f627f6b71b031ed502cf02b1b0eb8bb2dfd72e2e354
-  data.tar.gz: 26f65f4e16312694c1163b0de7fe0545ca8ad7b8e1cdea647d931f131de6aaef89f0fe8b39be72509932295c1ad1521066c88722d5654cb6a856891c1c19e860
+  metadata.gz: f413d3853e97f512b1a600503f47509f0afc0918a6dbe4be2b1efd9d23e4acc267d2eba6d82a7fdf6a6b2ac6408003c084a486c2cdc8d4ea4e0cf4f14651d59f
+  data.tar.gz: 599d5fdeab672da7018878e4a836d4fd9fdf76fb27e797208aabee9fb3a359ae988c42f39f1c09c17352c80d826ecf35445d7b0a245c8fc15af54675a783506c

data/Gemfile

@@ -77,7 +77,7 @@ gem 'rbvmomi2', '3.10.0'
 gem 'rdoc', '7.0.4'
 gem 'rest-client', '2.1.0'
 gem 'rex', '2.0.13'
-gem 'rmagick', '7.1.0'
+gem 'rmagick', '7.1.1'
 gem 'rqrcode', '3.2.0'
 gem 'rspec', '3.13.2'
 gem 'rtesseract', '3.1.4'

data/README.md

@@ -84,6 +84,11 @@ message bus:
 
 ![Swarm Multi-Agent](documentation/diagrams/swarm-multi-agent.svg)
 
+Long-running turns also show **executive task briefs** (not raw commands) via
+`TaskSummarizer`: one full plan on submit (`emit_plan!`), then per-batch
+`about_to` lines keyed by `tool_counts_phrase` + `intent_phrase` with
+`last_brief_fp` duplicate suppression. When recent turns keep hitting the iteration ceiling, the Loop tightens the remaining runway (lower `max_iters` on local engines, text-only tail, no counterfactual fork) so the agent still finishes instead of thrashing.
+
 Full pages: [How PWN Works](documentation/How-PWN-Works.md) ·
 [All data-flow diagrams](documentation/Diagrams.md)
 
@@ -101,11 +106,11 @@ The complete wiki lives in this repo at **[`documentation/Home.md`](documentatio
 | [Installation](documentation/Installation.md) | [Build a Driver](documentation/Drivers.md) | [Mistakes (neg-feedback)](documentation/Mistakes.md) | [WWW (21)](documentation/WWW.md) |
 | [General Usage](documentation/General-PWN-Usage.md) | | [Extrospection](documentation/Extrospection.md) | [SDR / Radio](documentation/SDR.md) |
 | [Configuration](documentation/Configuration.md) | | [Swarm (multi-agent)](documentation/Swarm.md) | [Hardware](documentation/Hardware.md) |
-| [Configuration](documentation/Configuration.md) | | [Sessions](documentation/Sessions.md) · [Cron](documentation/Cron.md) | [Reports](documentation/Reporting.md) |
-| [`~/.pwn/` Persistence](documentation/Persistence.md) | | | [BurpSuite](documentation/BurpSuite.md) · [NmapIt](documentation/NmapIt.md) |
-| **[All Diagrams](documentation/Diagrams.md)** | | | [Metasploit](documentation/Metasploit.md) · [Fuzzing](documentation/Fuzzing.md) |
-| [Troubleshooting](documentation/Troubleshooting.md) | | | [Hardware](documentation/Hardware.md) · [Blockchain](documentation/Blockchain.md) |
-| [Contributing](documentation/Contributing.md) | | | [Bounty](documentation/Bounty.md) · [FFI](documentation/FFI.md) · [Banner](documentation/Banner.md) |
+| [`~/.pwn/` Persistence](documentation/Persistence.md) | | [Sessions](documentation/Sessions.md) · [Cron](documentation/Cron.md) | [Reports](documentation/Reporting.md) |
+| **[All Diagrams](documentation/Diagrams.md)** (29) | | | [BurpSuite](documentation/BurpSuite.md) · [NmapIt](documentation/NmapIt.md) |
+| [Troubleshooting](documentation/Troubleshooting.md) | | | [Metasploit](documentation/Metasploit.md) · [Fuzzing](documentation/Fuzzing.md) |
+| [Contributing](documentation/Contributing.md) | | | [Blockchain](documentation/Blockchain.md) · [Bounty](documentation/Bounty.md) |
+| | | | [FFI](documentation/FFI.md) · [Banner](documentation/Banner.md) |
 
 Rebuild every SVG from its Graphviz source:
 `cd documentation/diagrams && ./build.sh`
@@ -202,7 +207,7 @@ and the local [Contributing](documentation/Contributing.md) page.
 ### **Module Documentation** ###
 
 **Primary:** [`documentation/Home.md`](documentation/Home.md) - the full local
-wiki with 30+ pages and 26 SVG data-flow diagrams.
+wiki with 30+ pages and 29 SVG data-flow diagrams.
 
 **API reference:** [rubydoc.info/gems/pwn](https://www.rubydoc.info/gems/pwn),
 or in-REPL: `PWN::Plugins::BurpSuite.help`, `show-source`, `ls`.

data/documentation/Agent-Tool-Registry.md

@@ -16,7 +16,7 @@ toolsets; the JSON-Schema for each tool is what the model actually sees.
 | `skills` | `skill_list` · `skill_view` · `skill_create` · `skill_add_reference` · `skill_delete` · `skill_migrate_legacy` | `~/.pwn/skills/<name>/SKILL.md` (**[agentskills.io](https://agentskills.io) spec**; legacy flat `*.md` auto-migrated) |
 | `sessions` | `sessions_list` · `sessions_view` · `sessions_current` · `sessions_delete` · `sessions_stats` | `PWN::Sessions` → `~/.pwn/sessions/` |
 | `learning` | `learning_note_outcome` · `learning_reflect` · `learning_distill_skill` · `learning_stats` · `learning_outcomes` · `learning_consolidate` · `learning_reset` · `learning_auto_introspect_toggle` · **`mistakes_list`** · **`mistakes_record`** · **`mistakes_resolve`** · **`mistakes_reset`** · **`reward_judge`** · **`reward_prm`** · **`reward_sentinel`** · **`reward_preferences`** · **`reward_export_dpo`** · **`reward_warm_sentinel`** · **`reward_scrub_preferences`** · **`reward_preference_balance`** · **`curriculum_practice`** · **`curriculum_train`** · **`curriculum_hindsight`** · **`curriculum_offline_judge`** · **`curriculum_preference_balance`** · **`learning_purge_noise`** | `PWN::AI::Agent::Learning` + `Mistakes` + `Reward` + `Curriculum` → `~/.pwn/learning.jsonl` + `~/.pwn/mistakes.json` + `~/.pwn/preferences.jsonl` + `~/.pwn/curriculum/` + `~/.pwn/finetune/` |
-| `reward` | **`reward_generator_mix`** | `PWN::AI::Agent::Reward.generator_mix` → W1 online source-mix controller (`preferences.jsonl`) |
+| `reward` | **`reward_generator_mix`** | `PWN::AI::Agent::Reward.generator_mix` → online preference source-mix controller (`preferences.jsonl`) |
 | `curriculum` | **`curriculum_practice_kpi`** | `PWN::AI::Agent::Curriculum.practice_kpi` → `~/.pwn/curriculum_kpi.jsonl` |
 | `metrics` | `metrics_summary` · `metrics_reset` | `PWN::AI::Agent::Metrics` → `~/.pwn/metrics.json` |
 | `extrospection` | `extro_snapshot` · `extro_drift` · `extro_observe` · `extro_observations` · `extro_intel` · **`extro_watch`** · **`extro_verify`** · **`extro_rf_tune`** · **`extro_osint`** · **`extro_serial`** · **`extro_telecomm`** · **`extro_packet`** · **`extro_vision`** · **`extro_voice`** · `extro_correlate` · `extro_stats` · `extro_reset` · `extro_auto_toggle` | `PWN::AI::Agent::Extrospection` (+ Serial/Packet/OCR/Voice/BareSIP/TransparentBrowser/GQRX) → `~/.pwn/extrospection.json` |

data/documentation/CLI-Drivers.md

@@ -1,6 +1,6 @@
 # CLI Drivers - `bin/pwn_*`
 
-53 headless executables (52 `pwn_*` + `pwn`), each a thin `OptionParser`
+54 headless executables (53 `pwn_*` + `pwn`), each a thin `OptionParser`
 wrapper over one plugin (or one workflow). They exist so CI/CD can call PWN
 without a REPL or an LLM.
 

data/documentation/Configuration.md

@@ -99,13 +99,17 @@ ai:
   agent:
     native_tools: true             # Use provider-native tool_calls / function-calling. false → legacy text-parsed tool protocol.
     max_iters: 25                  # Hard cap on tool-call rounds per user turn before a forced final answer.
+    task_summary: true             # Executive task briefs via TaskSummarizer (plan + about_to). false disables.
+    task_summary_every: 5          # When task_summary_verbose: emit Progress every N completed tools.
+    task_summary_interval_s: 8.0   # When verbose: also emit when this many seconds elapsed.
+    task_summary_verbose: false    # Mid-flight Progress/Finished lines (default: only plan + about_to).
     max_depth: 3                   # Recursion guard: how many levels deep agent_ask/agent_debate sub-agents may spawn sub-agents.
     auto_introspect: true          # Run Learning.auto_introspect (outcome logging + lesson mining) after every final answer.
     auto_extrospect: false         # Optional ambient baseline (host/repo/env ONLY - never launches burpsuite/zaproxy/msf/gqrx). Sense tools (intel/verify/watch/rf_tune/observe) stay on-demand.
-    plan_first: ~                  # Plan-then-act pre-pass: force the model to externalise a numbered tool plan BEFORE its first dispatch. nil = auto (true when ai.active == ollama).
+    plan_first: ~                  # Plan-then-act pre-pass: force the model to externalize a numbered tool plan BEFORE its first dispatch. nil = auto (true when ai.active == ollama).
     tool_router: false             # Dynamic tool-set slimming: ship only Registry::CORE_TOOLS + top-K keyword-relevant schemas per turn (helps small models route correctly).
     escalation_persona: ~          # Swarm persona name to ask for a 3-line corrective hint once a local model burns ≥ Loop::ESCALATE_AFTER_FAILS in-turn failures. nil = disabled.
-    toolsets: ~                    # Allow-list of toolsets exposed to the agent. nil = all. Valid: cron, extrospection, learning, memory, metrics, pwn, sessions, skills, swarm, terminal.
+    toolsets: ~                    # Allow-list of toolsets exposed to the agent. nil = all. Valid: cron, curriculum, extrospection, learning, memory, metrics, pwn, reward, sessions, skills, swarm, terminal.
     extrospection:
       web:
         anchors:                   # URLs the headless browser fingerprints on extro_snapshot(sections:[:web]). Alias: web_anchors.
@@ -124,7 +128,7 @@ ai:
       osint:
         ttl: 86400                 # Observation TTL for :osint (default 1 day).
         proxy: ~                   # Optional upstream proxy for OSINT HTTP feeds.
-        api_keys:                  # Per-feed API keys (all optional; keyed feeds return {skipped:} when absent). ENV fallbacks also honoured.
+        api_keys:                  # Per-feed API keys (all optional; keyed feeds return {skipped:} when absent). ENV fallbacks also honored.
           shodan: ...              #   ← SHODAN_API_KEY
           hunter: ...              #   ← HUNTER_API_KEY
           abuseipdb: ...           #   ← ABUSEIPDB_API_KEY
@@ -224,7 +228,7 @@ PWN::Config.refresh_env
 
 ---
 
-## Exhaustive key reference
+## Full key reference
 
 ### `ai` - AI engines & agent loop
 
@@ -258,11 +262,15 @@ PWN::Config.refresh_env
 | Key path | Type | Default | Consumed by | Purpose |
 |---|---|---|---|---|
 | `ai.agent.native_tools` | Boolean | `true` | `PWN::Plugins::REPL` (`pwn-ai` cmd) | Use provider-native `tool_calls` / function-calling. `false` falls back to the legacy text-parsed tool protocol. |
-| `ai.agent.max_iters` | Integer | `25` | `PWN::AI::Agent::Loop.run`, `PWN::AI::Agent::Swarm` | Hard cap on tool-call rounds per user turn before a forced final answer. |
+| `ai.agent.max_iters` | Integer | `25` | `PWN::AI::Agent::Loop.run`, `PWN::AI::Agent::Swarm` | Hard cap on tool-call rounds per user turn before a forced final answer. When budget-exhaustion pressure is high, the effective cap is tightened (stricter on local/ollama than remote) so long multi-step goals keep a usable runway without thrashing. |
+| `ai.agent.task_summary` | Boolean | `true` | `PWN::AI::Agent::TaskSummarizer`, `Loop` | Master switch for executive task briefs (`emit_plan!` / `about_to`). |
+| `ai.agent.task_summary_every` | Integer | `5` | `TaskSummarizer.every_n` | Verbose progress cadence (tools). |
+| `ai.agent.task_summary_interval_s` | Float | `8.0` | `TaskSummarizer.interval_s` | Verbose progress cadence (seconds). |
+| `ai.agent.task_summary_verbose` | Boolean | `false` | `TaskSummarizer.verbose?` | Emit mid-flight `Progress:` / `Finished:` lines; default keeps only plan + about_to. |
 | `ai.agent.max_depth` | Integer | `3` | `PWN::AI::Agent::Swarm` | Recursion guard for `agent_ask` / `agent_debate` sub-agents spawning sub-agents. |
 | `ai.agent.auto_introspect` | Boolean | `true` | `PWN::AI::Agent::Learning.auto_introspect` | Run outcome logging + lesson mining after every final answer. Toggle live via `learning_auto_introspect_toggle`. |
 | `ai.agent.auto_extrospect` | Boolean | `false` | `PWN::AI::Agent::Extrospection.auto_extrospect` | Optional ambient baseline after every final answer (`AUTO_SECTIONS` = host/repo/env only; never spawns GUI/JVM tools). Prefer on-demand sense tools (`intel`/`verify`/`watch`/`rf_tune`/`observe`). Toggle live via `extro_auto_toggle`. |
-| `ai.agent.toolsets` | Array\<String\> \| `nil` | `nil` (all) | `bin/pwn`, `PWN::Plugins::REPL`, `PWN::AI::Agent::Registry` | Allow-list of toolsets exposed to the agent. Valid: `cron`, `extrospection`, `learning`, `memory`, `metrics`, `pwn`, `sessions`, `skills`, `swarm`, `terminal`. |
+| `ai.agent.toolsets` | Array\<String\> \| `nil` | `nil` (all) | `bin/pwn`, `PWN::Plugins::REPL`, `PWN::AI::Agent::Registry` | Allow-list of toolsets exposed to the agent. Valid: `cron`, `curriculum`, `extrospection`, `learning`, `memory`, `metrics`, `pwn`, `reward`, `sessions`, `skills`, `swarm`, `terminal`. |
 | `ai.agent.plan_first` | Boolean \| `nil` | `nil` (auto: `true` when `ai.active == ollama`) | `PWN::AI::Agent::Loop.plan_first` | Plan-then-act pre-pass: the model must emit a numbered tool plan (as an assistant message) *before* it may dispatch anything. Cheap chain-of-thought scaffolding for local models. |
 | `ai.agent.tool_router` | Boolean | `false` | `PWN::AI::Agent::Registry.definitions` | Dynamic tool-set slimming: expose only `Registry::CORE_TOOLS` + the top-K keyword-relevant schemas for *this* request. Ties break on historical `Metrics` success rate so the router itself is a learned component. |
 | `ai.agent.escalation_persona` | String \| `nil` | `nil` | `PWN::AI::Agent::Loop.escalate` → `Swarm.ask` | Circuit-breaker: once a local model accumulates ≥ `Loop::ESCALATE_AFTER_FAILS` in-turn failures, ask this Swarm persona for a 3-line corrective hint (injected as a synthetic tool result). The local model still authors the final answer so Learning/Metrics stay attributed. |
@@ -283,7 +291,7 @@ PWN::Config.refresh_env
 | `ai.agent.extrospection.rf.ttl` | Integer | `300` | `Extrospection.rf_tune` | TTL (seconds) for `:rf` observations written by `extro_rf_tune` (ephemeral radio content). |
 | `ai.agent.extrospection.osint.ttl` | Integer | `86400` | `Extrospection.osint` | TTL (seconds) for `:osint` observations written by `extro_osint`. |
 | `ai.agent.extrospection.osint.proxy` | String | - | `Extrospection.osint` | Optional upstream proxy for OSINT HTTP feeds. |
-| `ai.agent.extrospection.osint.api_keys.<feed>` | String | - | `Extrospection.osint_api_keys` | Per-feed API keys for keyed OSINT sources: `shodan`, `hunter`, `abuseipdb`, `virustotal`, `greynoise`, `haveibeenpwned`, `securitytrails`, `steam`. Keyed feeds return `{skipped:}` when absent. ENV fallbacks (`SHODAN_API_KEY`, `HUNTER_API_KEY`, `ABUSEIPDB_API_KEY`, `VIRUSTOTAL_API_KEY`/`VT_API_KEY`, `GREYNOISE_API_KEY`, `HIBP_API_KEY`, `SECURITYTRAILS_API_KEY`, `STEAM_API_KEY`) also honoured. Redacted. |
+| `ai.agent.extrospection.osint.api_keys.<feed>` | String | - | `Extrospection.osint_api_keys` | Per-feed API keys for keyed OSINT sources: `shodan`, `hunter`, `abuseipdb`, `virustotal`, `greynoise`, `haveibeenpwned`, `securitytrails`, `steam`. Keyed feeds return `{skipped:}` when absent. ENV fallbacks (`SHODAN_API_KEY`, `HUNTER_API_KEY`, `ABUSEIPDB_API_KEY`, `VIRUSTOTAL_API_KEY`/`VT_API_KEY`, `GREYNOISE_API_KEY`, `HIBP_API_KEY`, `SECURITYTRAILS_API_KEY`, `STEAM_API_KEY`) also honored. Redacted. |
 | `ai.agent.extrospection.osint.social.sites_file` | Path | `etc/osint/social_sites.json` | `Extrospection.osint_social_sweep` | JSON of `{sites:{Name:{url:"...{u}...",absent_status:[404],absent_body:[...],head:bool}}}` used by the `:social_sweep` presence check. Vendored subset of sherlock-project (MIT). Override to add/remove platforms. |
 | `ai.agent.extrospection.osint.social.max_threads` | Integer | `16` | `Extrospection.osint_social_sweep` | Concurrency for the presence sweep (`Concurrent::FixedThreadPool`). |
 | `ai.agent.extrospection.osint.social.max_sites` | Integer | `120` | `Extrospection.osint_social_sweep` | Hard cap on sites loaded from `sites_file`. |

data/documentation/Contributing.md

@@ -9,7 +9,7 @@ lib/pwn/migrate.rb      # PWN::Migrate - ~/.pwn state doctor / auto-migrator
 lib/pwn/plugins/        # 66 plugin modules
 lib/pwn/ai/agent/       # agent core
 lib/pwn/ai/agent/tools/ # LLM tool registrations
-bin/                    # 52 pwn_* drivers + pwn (incl. pwn_setup)
+bin/                    # 53 pwn_* drivers + pwn (incl. pwn_setup)
 spec/                   # RSpec (incl. conventions_spec)
 documentation/          # this wiki + diagrams
 ```

data/documentation/Cron.md

@@ -37,14 +37,14 @@ invokes `PWN::Cron.run(<id>)` on schedule.
 | `learning_consolidate_nightly` | `0 5 * * *` | `PWN::AI::Agent::Learning.consolidate` |
 
 - **practice** - top unresolved `Mistakes` under `Reward.judge`; auto-`resolve` with ≥2 holdouts
-- **offline_judge** - P3 backfill of ORM/PRM labels + W3 calibration from PLAN `p(success)=` so `:failure_only` local introspect does not starve the corpus; also runs **P10** `Reward.warm_sentinel` so the R3 ring can fill on local hosts
-- **dedupe (P13)** - legacy alias `offline_judge_nightly` counts as the same job; `install_defaults` will not double-seed `30 3 * * *` and disables the alias when both exist
+- **offline_judge** - backfill outcome/process labels + plan calibration from PLAN `p(success)=` so `:failure_only` local introspect does not starve the corpus; also runs `Reward.warm_sentinel` so the reward-sentinel window can fill on local hosts
+- **dedupe** - legacy alias `offline_judge_nightly` counts as the same job; `install_defaults` will not double-seed `30 3 * * *` and disables the alias when both exist
 - **train** - export SFT + balanced DPO, LoRA-train `pwn-vN+1`, replay Mistakes.top, promote only on win. `dry_run: false` only with a trainer+GPU
-- **consolidate** - M1/M3 memory GC so the injected MEMORY block stays high-signal
+- **consolidate** - memory GC (semantic merge + importance eviction) so the injected MEMORY block stays high-signal
 
 See [Reinforcement Learning](Reinforcement-Learning.md).
 
-`cron_disable(id:)` turns either off; `install_defaults` is idempotent (including P13 alias collapse) and
+`cron_disable(id:)` turns either off; `install_defaults` is idempotent (including legacy alias collapse) and
 never overwrites a job you already have with the same name.
 
 ## Example

data/documentation/Diagrams.md

@@ -1,6 +1,6 @@
 # PWN Data-Flow Diagrams
 
-28 SVG diagrams, all rendered from Graphviz sources in
+29 SVG diagrams, all rendered from Graphviz sources in
 [`diagrams/dot/`](diagrams/dot/) with a single shared visual theme
 (see [`_THEME.md`](diagrams/dot/_THEME.md)). Rebuild everything with:
 
@@ -51,6 +51,11 @@ groups) so lines never criss-cross.
 [source](diagrams/dot/pwn-ai-feedback-learning-loop.dot) · doc: [Skills, Memory & Learning](Skills-Memory-Learning.md)
 ![pwn-ai-feedback-learning-loop](diagrams/pwn-ai-feedback-learning-loop.svg)
 
+
+### TaskSummarizer - Executive Task Briefs
+[source](diagrams/dot/task-summarizer.dot) · doc: [pwn-ai Agent](pwn-ai-Agent.md)
+![task-summarizer](diagrams/task-summarizer.svg)
+
 ### Mistakes - Negative-Feedback Loop
 [source](diagrams/dot/mistakes-negative-feedback.dot) · doc: [Mistakes](Mistakes.md)
 ![mistakes-negative-feedback](diagrams/mistakes-negative-feedback.svg)
@@ -63,7 +68,7 @@ groups) so lines never criss-cross.
 [source](diagrams/dot/ai-integration-tool-calling.dot) · doc: [AI Integration](AI-Integration.md)
 ![ai-integration-tool-calling](diagrams/ai-integration-tool-calling.svg)
 
-### Agent Tool Registry (10 toolsets · 71 tools)
+### Agent Tool Registry (12 toolsets · 78 tools)
 [source](diagrams/dot/agent-tool-registry.dot) · doc: [Agent Tool Registry](Agent-Tool-Registry.md)
 ![agent-tool-registry](diagrams/agent-tool-registry.svg)
 

data/documentation/Drivers.md

@@ -1,7 +1,7 @@
 # Drivers - Turn a REPL Session into a Shipped Binary
 
 A **driver** is a small executable in `bin/` that wires `OptionParser` to one
-or more `PWN::` calls. All 52 shipped `pwn_*` binaries follow the same
+or more `PWN::` calls. All 53 shipped `pwn_*` binaries follow the same
 15-line template.
 
 ![History → Driver → CI](diagrams/history-to-drivers.svg)

data/documentation/Extrospection.md

@@ -453,7 +453,7 @@ cron_create(name: 'memory_revalidate', schedule: '0 4 * * 0',
 ```ruby
 PWN::Env[:ai][:agent][:extrospection][:web] = {
   anchors:          ['https://target.acme/api/version', ...],  # else DEFAULT_WEB_ANCHORS
-  proxy:            'tor',      # or 'http://127.0.0.1:8080' (Burp) - honoured by verify/watch/probe_web
+  proxy:            'tor',      # or 'http://127.0.0.1:8080' (Burp) - honored by verify/watch/probe_web
   max_anchors:      8,
   per_page_timeout: 15,
   screenshot:       false,      # → ~/.pwn/extrospection/web/<host>.png

data/documentation/Home.md

@@ -20,15 +20,15 @@
 | [Installation](Installation.md) | `gem install pwn` → `pwn setup` doctor/provisioner · capability profiles · `--migrate` state doctor |
 | [General Usage](General-PWN-Usage.md) | Day-one cheat sheet |
 | [Configuration](Configuration.md) | `~/.pwn/pwn.yaml` (encrypted) - engines, keys, agent options · `pwn-vault` |
-| **[All Data-Flow Diagrams](Diagrams.md)** | **28 SVGs** in one scrollable page |
+| **[All Data-Flow Diagrams](Diagrams.md)** | **29 SVGs** in one scrollable page |
 
 ## 🚪 Entry Points
 
 | | |
 |---|---|
 | [The `pwn` REPL](pwn-REPL.md) | Pry shell with the whole `PWN::` namespace pre-loaded |
-| [`pwn-ai` Autonomous Agent](pwn-ai-Agent.md) | Natural-language TUI + `pwn --ai PROMPT` one-shot |
-| [CLI Drivers `bin/pwn_*`](CLI-Drivers.md) | 53 headless executables for CI/CD |
+| [`pwn-ai` Autonomous Agent](pwn-ai-Agent.md) | Natural-language TUI + `pwn --ai PROMPT` one-shot · **TaskSummarizer** briefs · iteration budget guard |
+| [CLI Drivers `bin/pwn_*`](CLI-Drivers.md) | 53 `pwn_*` + `pwn` headless executables for CI/CD |
 | [Drivers (build your own)](Drivers.md) | Turn a REPL session into a shipped binary |
 
 ## 🤖 AI Subsystem  (`PWN::AI`)
@@ -39,7 +39,7 @@
 | [Agent Tool Registry](Agent-Tool-Registry.md) | 12 toolsets · **78** LLM-callable tools |
 | [Memory · Skills · Learning](Skills-Memory-Learning.md) | Introspection - the self-improvement loop |
 | [Mistakes](Mistakes.md) | **Negative feedback** - fingerprint failures · do-NOT-repeat · `[REPEATING]`/`[REGRESSED]` · inline self-correction |
-| [Reinforcement Learning](Reinforcement-Learning.md) | **`Reward` + `Curriculum`** - ORM/PRM · P9/P14 trajectory DPO · P15 scrub · P16 warm · P17 budget skill · P18 PRM→rank · P19 diet gate · export-ready LoRA |
+| [Reinforcement Learning](Reinforcement-Learning.md) | **`Reward` + `Curriculum`** - outcome/process judges · preference ledger · self-play practice · export-ready LoRA gate |
 | [Extrospection](Extrospection.md) | World-awareness - snapshot · drift · intel · **watch** · **verify** · **rf_tune** · **osint** · serial · telecomm · packet · vision · voice · correlate |
 | [Swarm (Multi-Agent)](Swarm.md) | Personas · ask · debate · broadcast · shared bus |
 | [Sessions](Sessions.md) | Transcript persistence + reflection |

data/documentation/How-PWN-Works.md

@@ -20,21 +20,22 @@ hardware).
 | `pwn-ai` | `lib/pwn/ai/agent/loop.rb` | Agent TUI inside the REPL |
 | `pwn --ai PROMPT` | `bin/pwn` | Headless one-shot agent (CI-friendly) |
 | `pwn setup` | `lib/pwn/setup.rb` · `bin/pwn_setup` | Post-install doctor + capability provisioner + `--migrate` state doctor (also `pwn --setup[=PROFILE]`) |
-| `bin/pwn_*` | 52 files | Thin OptionParser wrappers over one plugin each |
+| `bin/pwn_*` | 53 files | Thin OptionParser wrappers over one plugin each |
 | `PWN::Cron` | `lib/pwn/cron.rb` | Scheduled jobs → any of the above (nightly self-play + weekly weight-loop seeded by default) |
 
 ## L2 - AI agent core  (`lib/pwn/ai/agent/`)
 
 | Module | Role |
 |---|---|
-| `Loop` | plan → dispatch tool_calls → observe → repeat until final answer |
+| `Loop` | plan → **TaskSummarizer** briefs → dispatch tool_calls → observe → repeat until final answer; tightens runway when recent turns exhausted the budget |
+| **`TaskSummarizer`** | Executive UX: `emit_plan!` (full goal once) · `about_to` per tool batch (`tool_counts_phrase` + `intent_phrase`) · `last_brief_fp` dedup · plan_idx advance |
 | `Registry` | JSON-Schema function definitions grouped into 12 **toolsets** · **78 tools** |
 | `Dispatch` / `Result` | execute a tool, capture stdout/value/error/duration |
 | `PromptBuilder` | inject MEMORY / SKILLS / LEARNING / **KNOWN MISTAKES + FIXES** / METRICS / EXTROSPECTION blocks |
 | `Metrics` · `Learning` · `Reflect` | **introspection** - how well am I doing? |
 | `Mistakes` | **negative feedback** - fingerprint failures, do NOT repeat, `[REPEATING]`/`[REGRESSED]`, inline `correction_hint` |
-| **`Reward`** | **R1** ORM `judge` · **R2** PRM per-step credit · **R3** `sentinel` reward-hacking detector · **R4** `semantic_ok` · **W1** DPO `preferences.jsonl` |
-| **`Curriculum`** | **S1** mistake-driven self-play `practice` · **S2** `counterfactual` A/B · **S3** tool-armed `critic` · **S4** `red_team_plan` · **C3** `hindsight` (HER) · **W2** `train_and_gate` regression-gated LoRA |
+| **`Reward`** | outcome `judge` · per-step process credit · `sentinel` (proxy vs judge drift) · `semantic_ok` · DPO `preferences.jsonl` |
+| **`Curriculum`** | mistake-driven self-play `practice` · `counterfactual` A/B · tool-armed `critic` · `red_team_plan` · `hindsight` (HER) · `train_and_gate` regression-gated LoRA |
 | `Extrospection` | **extrospection** - on-demand world sensing (`intel` · **`verify`** · **`watch`** · **`rf_tune`** · **`osint`** · `serial` · `telecomm` · `packet` · `vision` · `voice`) + ambient baseline (host · net · toolchain · repo · env · **rf** · **web**) joined to introspection via `correlate` |
 | `Swarm` | multi-agent personas over a shared JSONL bus |
 

data/documentation/Mistakes.md

@@ -100,7 +100,7 @@ upgraded", "the HackRF was unplugged", or "the target DOM moved". See
 [Extrospection](Extrospection.md#extro_correlate--the-point-of-the-whole-thing).
 
 
-## Preference pairs from resolve (W1 / P9)
+## Preference pairs from resolve
 
 `Mistakes.resolve` writes a DPO preference into `~/.pwn/preferences.jsonl`:
 
@@ -109,7 +109,7 @@ upgraded", "the HackRF was unplugged", or "the target DOM moved". See
   present; falls back to `AVOID_ARGS` / fix prose only when no trace exists
 
 Curriculum auto-resolve always attaches a winning tool trace so the ledger
-learns **trajectories**, not commentary. Write-time source quotas (P9) still
+learns **trajectories**, not commentary. Write-time source quotas still
 apply to pure prose resolves so resolve-monoculture cannot flood the ledger.
 
 **See also:** [Skills, Memory & Learning](Skills-Memory-Learning.md) ·

data/documentation/Persistence.md

@@ -16,7 +16,7 @@ Every byte PWN remembers between processes lives here.
 | **`preferences.jsonl`** | **`PWN::AI::Agent::Reward`** | JSON-per-line `{prompt,rejected,chosen,source}` | `rm` | **DPO/KTO/ORPO preference-pair ledger - user_correction · mistakes_resolve · counterfactual · critic · curriculum** |
 | **`mistakes.json`** | **`PWN::AI::Agent::Mistakes`** | **JSON `{sig → entry}`** | **`mistakes_reset`** | **failure fingerprints · cross-session count · fix · `[REPEATING]` · `[REGRESSED]`** |
 | `metrics.json` | `PWN::AI::Agent::Metrics` | JSON | `metrics_reset` | per-tool calls · success · avg_duration · last_error · **per-engine** sub-buckets · calibration |
-| `reward_sentinel.json` | `PWN::AI::Agent::Reward` | JSON | `rm` | R3 - proxy vs judge vs user-correction gap history |
+| `reward_sentinel.json` | `PWN::AI::Agent::Reward` | JSON | `rm` | proxy vs judge vs user-correction gap history |
 | `extrospection.json` | `PWN::AI::Agent::Extrospection` | JSON | `extro_reset` | host/net/toolchain/repo/env/**rf**/**web**/osint/serial/telecomm/packet/vision/voice snapshot + previous baseline + observations[] |
 | `extrospection/web/*.png` | `Extrospection` | PNG | `rm -rf` | headless-browser screenshots from `probe_web` / `extro_watch` (opt-in) |
 | `extrospection/packet/*.pcap` | `Extrospection` | pcap | `rm -rf` | bounded captures from `extro_packet(action: :capture)` |
@@ -27,12 +27,22 @@ Every byte PWN remembers between processes lives here.
 | `agents.yml` | `PWN::AI::Agent::Swarm` | YAML | edit / `agent_spawn` | persona registry |
 | `swarm/<id>/bus.jsonl` | `Swarm` | JSON-per-line | rm -rf | append-only multi-agent chat |
 | `swarm/<id>/personas.json` | `Swarm` | JSON | rm | persona → session_id map |
-| `curriculum/` | `PWN::AI::Agent::Curriculum` | JSONL | rm -rf | S1 self-play reproducers · W2 A/B eval sets |
+| `curriculum/` | `PWN::AI::Agent::Curriculum` | JSONL | rm -rf | self-play reproducers · train/gate A/B eval sets |
 | `finetune/*.jsonl` | `Learning.export_finetune` · `Reward.export_dpo` | ShareGPT / OpenAI / DPO JSONL | `rm` | supervised + preference datasets - feed to a LoRA over the local model |
 | `backup/<ts>/` | `PWN::Migrate` | tree copy | rm -rf | timestamped snapshot taken before every `--migrate --fix` |
 | `quarantine/` | `PWN::Migrate` | quarantined originals | rm -rf | corrupt/incompatible state files moved aside so the owner re-seeds |
 | `~/.pwn_history` | Pry | text | rm | REPL input history |
 
+
+## `memory.json` safety
+
+`PWN::Memory.save` **refuses** to overwrite a non-empty `memory.json` with `{}`
+unless you pass `force: true`. That blocks the failure mode where a parse error
+was rescued into an empty hash and then written back (full wipe). `memory_clear`
+and deleting the last key use `force: true` on purpose. Load no longer falls
+back to `{}` on parse failure for a non-empty file; it warns and keeps the bytes
+on disk for repair.
+
 ## Verify / repair the whole tree
 
 ```bash