tools tier

`tools-prompt-shaper`

prompt-shape: screen, correct, and structure operator input into a shaped prompt before it is sent to any model — never sends, only prints.

prompt-shape: screen, correct, and structure operator input into a shaped prompt before it is sent to any model — never sends, only prints.

Tier	tools
Role	unclassified (baselined)
Path	`crates/tools/prompt-shaper`
Edition	2021
Targets	prompt-shape, tools_prompt_shaper
Public items	19 across 4 modules
Tests	40

What it is for

tools-prompt-shaper — screen, correct, and structure operator input into a shaped prompt before it is sent to any model.

Pipeline order is the design (Sprint 3.21, Gate 1.5.6): hard-rules DLP screening runs FIRST so a credential is redacted before the local model — or any model — ever sees the text. The pipeline never sends anything anywhere; it prints a shaped prompt for the operator to review.

Modules land one per TDD task:

T2 screen — hard-rules DLP screening + redaction.
T3 model — the local-model typed ask, with degradation when the

model is unreachable or returns malformed output.

T4 template — deterministic template fill; corrections are applied

as spans, never a rewritten body.

T5 route and config — lookup routing and config plumbing.

T1 lands the crate skeleton and CLI surface only; modules are added as their task lands.

Capabilities

`config (other)`

T5: config plumbing. SHAPER_MODEL, OLLAMA_URL, and a hard-rules-only

Item
`pub const DEFAULT_OLLAMA_URL: & str`
`fn parse_env_file(contents : & str) -> HashMap <String, String>`

`Config`

T5: config plumbing. SHAPER_MODEL, OLLAMA_URL, and a hard-rules-only

Item
`pub struct Config`
`fn resolve(cli_model : Option <& str>, cli_ollama_url : Option <& str>, cli_no_model : bool, process_env : & HashMap <String, String>, file_env : & HashMap <String, String>, default_model : & str,) -> Config`

`model (other)`

T3: the local-model typed ask, and the degradation path when the model

Item
`pub const DEFAULT_MODEL: & str`
`pub const SYSTEM_PROMPT: & str`
`async fn ask(provider : & dyn AiProvider, model : & str, redacted_text : & str,) -> Result <ModelAsk, Degraded>`

`Degraded`

T3: the local-model typed ask, and the degradation path when the model

Item
`pub enum Degraded`
`Degraded :: fn message(& self) -> String`

`Intent`

T3: the local-model typed ask, and the degradation path when the model

Item
`pub enum Intent`

`ModelAsk`

T3: the local-model typed ask, and the degradation path when the model

Item
`pub struct ModelAsk`

`TypoCorrection`

T3: the local-model typed ask, and the degradation path when the model

Item
`pub struct TypoCorrection`

`ScreenResult`

T2: hard-rules DLP screening. Runs FIRST, before any model — local or

Item
`pub struct ScreenResult`
`ScreenResult :: fn has_findings(& self) -> bool`
`fn screen(text : & str) -> ScreenResult`

`template (other)`

T4: deterministic template fill.

Item
`fn fill(intent : Option <Intent>, text : & str, obfuscation_suspected : bool) -> String`

`AppliedCorrections`

T4: deterministic template fill.

Item
`pub struct AppliedCorrections`
`fn apply_corrections(text : & str, corrections : & TypoCorrection) -> AppliedCorrections`

`DroppedCorrection`

T4: deterministic template fill.

Item
`pub struct DroppedCorrection`

How to use it

No examples/ target and no doctest in this crate's rustdoc. The tests listed under Verification are the closest executable usage.

Module structure

tools_prompt_shaper

config
model
screen
template

flowchart TD
  n_tools_prompt_shaper["tools_prompt_shaper"]
  n_tools_prompt_shaper --> n_config["config"]
  n_tools_prompt_shaper --> n_model["model"]
  n_tools_prompt_shaper --> n_screen["screen"]
  n_tools_prompt_shaper --> n_template["template"]

Public surface

`config`

Item	What it is
`pub const DEFAULT_OLLAMA_URL: & str`	Default Ollama base URL, matching `infrastructure_ai::OllamaProvider::local`.
`pub struct Config`	Resolved configuration for one run.
`fn parse_env_file(contents : & str) -> HashMap <String, String>`	Parse a `KEY=VALUE` config file in the `~/.config/claude-bridge.env` convention: blank lines and `#`-prefixed comment lines are ignored, no quoting or escaping
`fn resolve(cli_model : Option <& str>, cli_ollama_url : Option <& str>, cli_no_model : bool, process_env : & HashMap <String, String>, file_env : & HashMap <String, String>, default_model : & str,) -> Config`	Resolve config from explicit sources

`model`

Item	What it is
`pub const DEFAULT_MODEL: & str`	Default local model (Sprint 3.21): general reasoning, not a coder model — intent classification and orthography are not code-shaped tasks.
`pub const SYSTEM_PROMPT: & str`	System prompt for the single typed-ask call
`pub enum Intent`	The intent the model assigns to the (already-screened) text
`pub struct TypoCorrection`	One correction as a span-and-replacement pair — never a rewritten body
`pub struct ModelAsk`	The three typed asks the local model answers in one call.
`pub enum Degraded`	Why the model step did not produce a `ModelAsk`
`Degraded :: fn message(& self) -> String`	One-line, operator-facing explanation of what was unavailable and why
`async fn ask(provider : & dyn AiProvider, model : & str, redacted_text : & str,) -> Result <ModelAsk, Degraded>`	Ask the local model to classify intent, propose typo corrections, and flag suspected obfuscation over `redacted_text`

`screen`

Item	What it is
`pub struct ScreenResult`	Result of the hard-rules screening pass.
`ScreenResult :: fn has_findings(& self) -> bool`	Whether screening found anything to redact.
`fn screen(text : & str) -> ScreenResult`	Screen `text` with `dlp-detect`'s zero-config default policy and return the redacted text plus the findings that drove the redaction.

`template`

Item	What it is
`pub struct AppliedCorrections`	Result of splicing corrections into text: which ones landed, and which were dropped (and why they don't appear in `text`).
`pub struct DroppedCorrection`	A correction that was not applied, and why.
`fn apply_corrections(text : & str, corrections : & TypoCorrection) -> AppliedCorrections`	Splice each correction's `replacement` in place of the first exact, non-fenced occurrence of its `span`
`fn fill(intent : Option <Intent>, text : & str, obfuscation_suspected : bool) -> String`	Fill the deterministic skeleton for `intent` around `text`

No pub use re-exports: every item above is declared in this crate.

Boundary

Reaches into foundation, infrastructure.

Shares tier tools with 84 other crates: tools-advisory-reach, tools-archive-guard, tools-artifact-scaffold, tools-ask-ai-core, tools-ask-ais, tools-ask-gemini, tools-book, tools-book-report, … (84 total).

_What this crate deliberately does NOT own is a judgment. No committed registry records one for it, so none is stated here._

Where it sits

Tier (ontology)	`tools`
Architectural role (taxonomy)	`unclassified (baselined)`
Location	`crates/tools/prompt-shaper`
Vocabulary in force (lexicon)	`current`

Tier flow. Which tiers this crate's own edges cross.

flowchart LR
  n_tools["tools"] --> n_foundation["foundation"]
  n_tools["tools"] --> n_infrastructure["infrastructure"]

Dependencies

Runtime, in this workspace.

Crate	Tier	Optional	Only on
`foundation-bounded-io`	foundation	no	always
`infrastructure-ai`	infrastructure	no	always
`infrastructure-dlp-detect`	infrastructure	no	always

Runtime, from outside the workspace.

Crate	Requirement	Features	Optional	Only on
`clap`	`^4`	`derive`	no	always
`serde`	`^1`	`derive`, `derive`	no	always
`serde_json`	`^1`	—	no	always
`tokio`	`^1`	`full`, `macros`, `rt-multi-thread`	no	always

Development, in this workspace.

Crate	Tier	Optional	Only on
`tools-cli-conformance`	tools	no	always

Development, from outside the workspace.

Crate	Requirement	Features	Optional	Only on
`async-trait`	`^0.1`	—	no	always

Build. None.

Depended on by. Nothing in this workspace.

Signal flow — what reaches this crate, and what it reaches.

flowchart LR
  SELF["tools-prompt-shaper"]
  SELF -->|development| n_tools_cli_conformance["tools-cli-conformance"]
  SELF -->|runtime| n_foundation_bounded_io["foundation-bounded-io"]
  SELF -->|runtime| n_infrastructure_ai["infrastructure-ai"]
  SELF -->|runtime| n_infrastructure_dlp_detect["infrastructure-dlp-detect"]
  classDef self fill:#1f883d,stroke:#1f883d,color:#fff;
  class SELF self;

Feature flags

No Cargo features are defined: every capability is unconditional, so no consumer can receive a half-wired crate.

Targets

Kind	Name	Source
bin	`prompt-shape`	`src/main.rs`
lib	`tools_prompt_shaper`	`src/lib.rs`

Error model

No public error type was detected: no public item declares a type named *Error, and no public signature returns one.

Operational characteristics

Property	Evidence
async public surface	yes
async runtime	yes
database access	none detected
network I/O	none detected
unsafe code	none detected
environment variables	yes

No unsafe block, unsafe fn, unsafe impl or unsafe trait was found by the parser anywhere in this crate's source.

Configuration

Variable	Read in
`HOME`	`src/main.rs`

No workspace crate depends on this one.

Verification

Kind	Count
Unit tests	40
Integration tests	0
Examples	0
Doctests	0

Evidence by module. How often each public module is named by something executable.

Module	Tests	Examples	Consumers
`config`	4	0	0
`model`	7	0	0
`screen`	2	0	0
`template`	4	0	0

What the tests establish, by name:

all_defaults_when_nothing_is_set — src/config.rs
cli_flag_overrides_everything — src/config.rs
cli_no_model_flag_sets_skip_model_regardless_of_env — src/config.rs
file_env_is_used_when_process_env_absent — src/config.rs
ollama_url_precedence_matches_model_precedence — src/config.rs
parse_env_file_empty_input_yields_empty_map — src/config.rs
parse_env_file_skips_comments_and_blank_lines — src/config.rs
process_env_overrides_file_env — src/config.rs
shaper_no_model_env_var_false_or_zero_does_not_skip — src/config.rs
shaper_no_model_env_var_sets_skip_model — src/config.rs
a_correction_with_no_match_is_reported_as_dropped — src/main.rs
cli_definition_conforms — src/main.rs
findings_are_reported_and_redacted_in_the_body — src/main.rs
provider_failure_degrades_the_whole_pipeline_not_just_the_model_step — src/main.rs
skip_model_bypasses_the_provider_and_notes_it — src/main.rs
truncation_is_reported_when_flagged — src/main.rs
empty_typo_corrections_defaults_when_omitted — src/model.rs
malformed_json_degrades_without_panicking — src/model.rs
ollama_dead_port_degrades_without_failing_the_tool — src/model.rs
parses_a_well_formed_response — src/model.rs
provider_error_degrades_as_unreachable_not_a_failure — src/model.rs
raw_credential_never_reaches_the_model_call — src/model.rs
sends_the_model_and_json_mode_flag — src/model.rs
clean_input_passes_through_unchanged — src/screen.rs
empty_input_has_no_findings_and_is_unchanged — src/screen.rs
findings_carry_only_masked_values_never_raw — src/screen.rs
raw_value_never_survives_into_redacted_text_across_all_classes — src/screen.rs
ssn_is_found_and_redacted — src/screen.rs
a_correction_inside_a_fenced_block_is_dropped_never_applied — src/template.rs
a_correction_outside_a_fence_still_applies_when_another_fence_is_present — src/template.rs
_… 10 more_

Documentation coverage

Measure	Documented	Total
Public items with rustdoc	19	19
Public modules with a `//!` block	4	4

pie showData
    title Public items with rustdoc
    "Documented" : 19
    "No rustdoc detected" : 0

Metrics

Metric	Value
Rust source files	6
Source lines	1217
Code lines	891
Public API items	19
Public modules	4
Tests	40
Examples	0
Cargo features	0
Direct runtime dependencies	7
Workspace reverse dependencies	0

pie showData
    title Public API by kind
    "constant" : 3
    "enum" : 2
    "function" : 6
    "method" : 2
    "struct" : 6

pie showData
    title Rust source composition
    "Code" : 891
    "Blank or comment" : 326

Generation

Rendered by tools-corpus corpus readme from repository evidence alone, renderer schema 2, lexicon current. No model, network service or database was consulted. Regenerate with tools-corpus corpus readme --write; verify with --check.

All tools · Manual