feat: wire prompt-injection defenses into LLM tools (Item 10) - #69
Merged
Conversation
mingjerli
force-pushed
the
feat/wire-path-validation
branch
from
July 22, 2026 12:36
1a40e75 to
e8e685c
Compare
…nk warning, harden passthrough test
mingjerli
force-pushed
the
feat/wire-prompt-sanitization
branch
from
July 22, 2026 12:45
a610938 to
0c3d97f
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on the Item 7 PR (#68). Applies the four prompt-injection defense layers at the LLM call sites, wiring in the existing (previously dead)
prompt_sanitization.py:<data>(descriptions),<schema>/<question>(SQL gen),<sql>(explain), each with a do-not-follow directive.max_lengthfor large schema); SQL expressions viasanitize_sql_for_prompt.LLMTool.call_llm_structured(system/user role split); used by the description and explain paths._validate_description_output→ existing rule-based fallback; generated SQL blocked on destructive ops, passed through (with warning) when sqlglot can't parse.All new tests exercise public entry points and are pinned so they fail if a defense is removed. Full suite: 1554 passed / 70 skipped / 2 xfailed.
A whole-branch review flagged two same-class holes on entry points outside this plan's scope —
from_dbt_models(unvalidated .sql reads) andtable.pytable-level descriptions (unsanitized). Neither is overclaimed in the CHANGELOG; tracking separately.Design: plans/clgraph-items-7-10-wiring-design.md