feat: content extractors for manifests and entry points

Adds an optional `extract` field to priority rules that transforms file
content before budget accounting. Instead of including a full Cargo.toml
or truncating package.json to its first two lines, extractors produce a
focused summary (just the dependencies, just the pub mod declarations)
that fits in the content budget and actually orients an agent.

Four built-in extractors:
- json_keys: extract named top-level keys from JSON
- toml_sections: extract named sections from TOML
- lines_matching: extract lines starting with given prefixes
- api_surface: first 5 lines + all pub/export declarations

Default rules auto-extract for:
- Cargo.toml -> dependencies, dev-dependencies
- package.json -> dependencies, devDependencies, scripts
- go.mod -> require and module lines
- src/lib.*, **/mod.rs -> public API surface

The extract field is optional and defaults to None, so existing configs
work unchanged. Users can add extract specs to any priority rule in
dirpack.toml.

New file: src/packer/extractors.rs (~330 lines incl. 14 unit tests)
Modified: src/config.rs (extract field + defaults), src/packer/mod.rs
(3-line integration), src/priority.rs (test fixes), CHANGELOG.md,
default_config.toml

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: rawwerks

CHANGELOG.md

@@ -17,6 +17,7 @@ Semantic Versioning.
 - `content_weight` per file category controls how aggressively each category's files are included in the content phase. Defaults: code/docs 1.0, config 0.2, build/data 0.0. Configurable in dirpack.toml via `[categories.<name>].content_weight`.
 - `test_content_weight` and `fixture_content_weight` in `[priority]` (both default 0.0). Test and fixture files get spine + signatures but never content body. Set to 1.0 to restore old behavior.
 - Import-only snippet detection: truncated snippets where >80% of lines are import/require/use statements are skipped automatically.
+- Content extractors: optional `extract` field on priority rules transforms file content before budget accounting. Four built-in extractors: `json_keys` (package.json deps/scripts), `toml_sections` (Cargo.toml dependencies), `lines_matching` (go.mod require blocks), `api_surface` (pub mod/use/fn declarations from entry points). Default rules auto-extract for common manifests and lib.rs files. Configurable per-pattern in dirpack.toml.
 
 ### Changed
 - Content phase skips files under 50 bytes (fully represented by spine) and snippets truncated to fewer than 3 lines (eliminates single-import/shebang junk). Among equal-priority files, larger files now sort first for content inclusion instead of smaller ones.

default_config.toml

@@ -45,6 +45,12 @@ priority = 30
 content_weight = 0.0     # never include raw data content
 
 # Priority rules - higher = more important, included first in budget
+# Optional "extract" field transforms content before budget accounting.
+# Built-in extractors:
+#   json_keys:key1,key2         - include only named top-level JSON keys
+#   toml_sections:sec1,sec2     - include only named TOML sections
+#   lines_matching:prefix1|pfx2 - include lines starting with any prefix
+#   api_surface                 - first 5 lines + all pub/export declarations
 [[priority_rules]]
 pattern = "README*"
 priority = 200
@@ -60,14 +66,17 @@ priority = 200
 [[priority_rules]]
 pattern = "Cargo.toml"
 priority = 150
+extract = "toml_sections:dependencies,dev-dependencies"
 
 [[priority_rules]]
 pattern = "package.json"
 priority = 150
+extract = "json_keys:dependencies,devDependencies,scripts"
 
 [[priority_rules]]
 pattern = "go.mod"
 priority = 150
+extract = "lines_matching:require |module "
 
 [[priority_rules]]
 pattern = "src/main.*"
@@ -76,10 +85,12 @@ priority = 140
 [[priority_rules]]
 pattern = "src/lib.*"
 priority = 140
+extract = "api_surface"
 
 [[priority_rules]]
 pattern = "**/mod.rs"
 priority = 130
+extract = "api_surface"
 
 [[priority_rules]]
 pattern = "**/*_test.*"

src/config.rs

@@ -253,6 +253,17 @@ impl Category {
 pub struct PriorityRule {
     pub pattern: String,
     pub priority: i32,
+    /// Optional content extractor spec. When set, the content phase
+    /// transforms this file's content through the named extractor before
+    /// budget accounting. Format: `"extractor_name:arg1,arg2"`.
+    ///
+    /// Built-in extractors:
+    /// - `json_keys:key1,key2` — include only named top-level JSON keys
+    /// - `toml_sections:sec1,sec2` — include only named TOML sections
+    /// - `lines_matching:prefix1|prefix2` — include lines starting with any prefix
+    /// - `api_surface` — first 5 lines + all pub/export declarations
+    #[serde(default)]
+    pub extract: Option<String>,
 }
 
 /// Configurable priority weight adjustments.
@@ -400,50 +411,62 @@ fn default_priority_rules() -> Vec<PriorityRule> {
         PriorityRule {
             pattern: "README*".to_string(),
             priority: 200,
+            extract: None,
         },
         PriorityRule {
             pattern: "AGENTS.md".to_string(),
             priority: 200,
+            extract: None,
         },
         PriorityRule {
             pattern: "CLAUDE.md".to_string(),
             priority: 200,
+            extract: None,
         },
         PriorityRule {
             pattern: "Cargo.toml".to_string(),
             priority: 150,
+            extract: Some("toml_sections:dependencies,dev-dependencies".to_string()),
         },
         PriorityRule {
             pattern: "package.json".to_string(),
             priority: 150,
+            extract: Some("json_keys:dependencies,devDependencies,scripts".to_string()),
         },
         PriorityRule {
             pattern: "go.mod".to_string(),
             priority: 150,
+            extract: Some("lines_matching:require |module ".to_string()),
         },
         PriorityRule {
             pattern: "src/main.*".to_string(),
             priority: 140,
+            extract: None,
         },
         PriorityRule {
             pattern: "src/lib.*".to_string(),
             priority: 140,
+            extract: Some("api_surface".to_string()),
         },
         PriorityRule {
             pattern: "**/mod.rs".to_string(),
             priority: 130,
+            extract: Some("api_surface".to_string()),
         },
         PriorityRule {
             pattern: "**/*_test.*".to_string(),
             priority: 50,
+            extract: None,
         },
         PriorityRule {
             pattern: "**/test_*".to_string(),
             priority: 50,
+            extract: None,
         },
         PriorityRule {
             pattern: "**/*.lock".to_string(),
             priority: 10,
+            extract: None,
         },
     ]
 }