feat(ask): add extra_instructions support for /ask (#2473) by rosspeili · Pull Request #2482 · The-PR-Agent/pr-agent

rosspeili · 2026-06-29T12:14:39Z

/ask didn’t support extra_instructions like /review and /improve. This adds it for PR-wide and line-level ask.

Set it in .pr_agent.toml or per comment, eg. to refuse certain questions (like “rate this PR 1–10”) and tell the user why.

Fixes #2473

qodo-free-for-open-source-projects · 2026-06-29T12:15:28Z

PR Summary by Qodo

Add extra_instructions support for /ask prompts (PR-wide and line-level)

✨ Enhancement 🧪 Tests 📝 Documentation ⚙️ Configuration changes 🕐 20-40 Minutes

AI Description

• Add configurable extra_instructions for /ask and line-level /ask prompts.
• Thread extra_instructions into Jinja prompt templates and tool variables.
• Document usage and add tests validating prompt rendering behavior.

Diagram

graph TD
  U(["User PR comment (/ask)"]) --> T["/ask tool"] --> R["Render Jinja prompt"] --> L{{"LLM"}}
  C[("pr_questions config") ] --> T
  P["Prompt templates"] --> R

  subgraph Legend
    direction LR
    _u(["User"]) ~~~ _p["Process"] ~~~ _cfg[("Config")] ~~~ _ext{{"External"}}
  end

High-Level Assessment

The following are alternative approaches to this PR:

1. Separate settings keys for PR-wide vs line-level ask

➕ Allows more granular control (different guardrails for code-line questions vs general PR questions).
➕ Avoids surprising behavior if line-level prompts need stricter/looser constraints.
➖ More config surface area and documentation overhead.
➖ Higher chance of misconfiguration / user confusion.

2. Shared prompt macro/include for the extra_instructions block

➕ Avoids duplicating the same conditional block across multiple prompt templates.
➕ Easier future edits to the wording/format of the extra instructions section.
➖ Requires introducing/standardizing a template include mechanism in the prompt-loading path (if not already present).
➖ Slightly more indirection for prompt authors.

Recommendation: The PR’s approach (single pr_questions.extra_instructions value injected into both PR-wide and line-level /ask system prompts) is a good default: minimal configuration, consistent behavior, and easy to override per comment via settings. If future users request different constraints for line-level vs PR-wide asks, consider splitting the config key; otherwise keep the simpler unified setting.

Files changed (7) +102 / -1

Enhancement (4) +21 / -1

pr_line_questions_prompts.tomlInject extra_instructions block into line-level /ask prompt +9/-0

Inject extra_instructions block into line-level /ask prompt
• Adds a Jinja conditional section that includes user-provided 'extra_instructions' in the system prompt when non-empty, keeping the prompt unchanged when empty.
pr_agent/settings/pr_line_questions_prompts.toml

pr_questions_prompts.tomlInject extra_instructions block into PR-wide /ask prompt +9/-0

Inject extra_instructions block into PR-wide /ask prompt
• Adds a Jinja conditional section to the system prompt to include 'extra_instructions' when provided, enabling custom guardrails and response shaping for '/ask'.
pr_agent/settings/pr_questions_prompts.toml

pr_line_questions.pyPass extra_instructions into line-level /ask template variables +2/-1

Pass extra_instructions into line-level /ask template variables
• Extends the variable payload used for prompt rendering to include 'extra_instructions' from 'get_settings().pr_questions.extra_instructions'.
pr_agent/tools/pr_line_questions.py

pr_questions.pyPass extra_instructions into PR-wide /ask template variables +1/-0

Pass extra_instructions into PR-wide /ask template variables
• Adds 'extra_instructions' to the variables used by the PR-wide questions tool so the prompt templates can render the optional extra instructions block.
pr_agent/tools/pr_questions.py

Tests (1) +48 / -0

test_pr_questions_helpers.pyAdd tests for extra_instructions prompt rendering +48/-0

Add tests for extra_instructions prompt rendering

• Introduces unit tests verifying that 'extra_instructions' text appears in the rendered system prompt when set and that the block is omitted when empty. Covers both PR-wide and line-level '/ask' prompt templates.

tests/unittest/test_pr_questions_helpers.py

Documentation (1) +32 / -0

ask.mdDocument /ask configuration including extra_instructions +32/-0

Document /ask configuration including extra_instructions

• Adds a new "Configuration options" section describing 'extra_instructions', 'enable_help_text', and 'use_conversation_history'. Includes examples for setting 'extra_instructions' in '.pr_agent.toml' and overriding it in a PR comment.

docs/docs/tools/ask.md

Other (1) +1 / -0

configuration.tomlAdd pr_questions.extra_instructions default setting +1/-0

Add pr_questions.extra_instructions default setting
• Introduces 'extra_instructions = ""' under the '[pr_questions]' section to provide a default empty value for '/ask' customization.
pr_agent/settings/configuration.toml

qodo-free-for-open-source-projects · 2026-06-29T12:16:37Z

Code Review by Qodo

🐞 Bugs (3) 📘 Rule violations (1) 📎 Requirement gaps (0) 📜 Skill insights (0)

Context used

✅ Tickets: 🎫 Support "extra_instructions" for /ask

1. ~~Line prompt forces answering always~~ ✓ Resolved 📎 Requirement gap ⛨ Security

Description

The line-level /ask system prompt still includes unconditional “must always answer” language that
can conflict with extra_instructions intended to require refusals for disallowed questions (e.g.,
PR rating requests). This direct prompt conflict can undermine policy enforcement by causing the
model to answer when it should refuse.

Code

pr_agent/settings/pr_line_questions_prompts.toml[R8-15]

+{%- if extra_instructions %}
+
+
+Extra instructions from the user:
+======
+{{ extra_instructions }}
+======
+{%- endif %}

Evidence

PR Compliance ID 6 requires extra_instructions to steer behavior, including reliably refusing
disallowed questions. The cited /ask system prompt includes an extra_instructions section, but
it is preceded by an unconditional requirement to always answer / not avoid answering, which creates
a direct conflict with refusal-oriented extra_instructions and can prevent refusal behavior from
being applied as required.

Add support for extra_instructions in /ask command
pr_agent/settings/pr_line_questions_prompts.toml[4-16]
pr_agent/settings/pr_questions_prompts.toml[4-15]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The `/ask` system prompt (including the line-level prompt) contains unconditional language requiring the model to always answer questions / not avoid answering, which can contradict `extra_instructions` that are meant to restrict behavior and require refusal for disallowed question types.
## Issue Context
PR Compliance ID 6 expects `extra_instructions` to be enforceable and to steer behavior, including refusing disallowed requests (e.g., “rate this PR 1–10”) with an explanation. Because the current `/ask` prompt includes “must answer” instructions alongside an `extra_instructions` block, the unconditional “always answer” requirement can override or conflict with refusal requirements, undermining compliance for line-level questions.
## Fix Focus Areas
- pr_agent/settings/pr_line_questions_prompts.toml[4-16]
- pr_agent/settings/pr_questions_prompts.toml[4-15]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

2. Extra instructions logged 🐞 Bug ⛨ Security

Description

PRQuestions.run() logs the entire pr_questions config dict; with the new
pr_questions.extra_instructions field, any text configured there will be included in debug log
artifacts.

Code

pr_agent/settings/configuration.toml[132]
+extra_instructions = ""

Evidence
The PR adds extra_instructions under [pr_questions], and the /ask implementation logs the entire
pr_questions section as debug artifacts, which will now include that new field's value.
pr_agent/settings/configuration.toml[129-133]
pr_agent/tools/pr_questions.py[54-58]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`PRQuestions.run()` logs `dict(get_settings().pr_questions)` as an artifact. After this PR, that dict includes `extra_instructions`, so any sensitive internal guidance placed there will be written to logs.
### Issue Context
This PR introduces `pr_questions.extra_instructions` to support policy-like instructions for `/ask`. Those instructions may be more sensitive than typical boolean/tuning settings.
### Fix Focus Areas
- pr_agent/tools/pr_questions.py[54-58]
- pr_agent/settings/configuration.toml[129-133]
### Implementation notes
- When building `relevant_configs`, omit `extra_instructions` entirely, or replace its value with a redacted form (e.g., `"<redacted>"` or `"<redacted: N chars>"`).
- Keep the rest of the config dump intact for debugging.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

3. No token budgeting line-ask 🐞 Bug ☼ Reliability

Description

PR_LineQuestions renders and sends prompts directly to the model without any token-budget
enforcement; with the newly-added extra_instructions block, long instructions can push the request
over the model context limit and fail the AI call.

Code

pr_agent/tools/pr_line_questions.py[R38-39]

+            "conversation_history": "",
+            "extra_instructions": get_settings().pr_questions.extra_instructions,

Evidence
The PR adds extra_instructions into the line-level system prompt and passes it via self.vars,
but PR_LineQuestions._get_prediction() calls the model with fully-rendered prompts and does not
use any token-budgeting/clipping logic before chat_completion.
pr_agent/settings/pr_line_questions_prompts.toml[8-16]
pr_agent/tools/pr_line_questions.py[31-45]
pr_agent/tools/pr_line_questions.py[152-167]
pr_agent/tools/pr_questions.py[40-43]
pr_agent/algo/pr_processing.py[67-96]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`PR_LineQuestions._get_prediction()` sends `system_prompt`/`user_prompt` to `chat_completion` without any clipping/token budgeting. This PR adds `extra_instructions` into the line-question system prompt, increasing the chance the combined prompt exceeds the model context window.
### Issue Context
Unlike `PRQuestions`, the line-level flow does not use `get_pr_diff(...)` or any other pruning step before calling the model.
### Fix Focus Areas
- pr_agent/tools/pr_line_questions.py[31-45]
- pr_agent/tools/pr_line_questions.py[152-167]
- pr_agent/settings/pr_line_questions_prompts.toml[1-20]
### Implementation notes
Implement a pre-flight token budget check in `PR_LineQuestions._get_prediction()` (or just before it calls `chat_completion`):
- Count tokens for `system_prompt` + `user_prompt`.
- If over limit, clip least-critical fields first (e.g., `conversation_history`, then `full_hunk`, and/or cap `extra_instructions` length) using existing clipping utilities.
- Ensure the truncation is deterministic and logged (without dumping full sensitive content).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

4. extra_instructions TOML formatting inconsistent 📘 Rule violation ⚙ Maintainability

Description

In pr_agent/settings/configuration.toml the new pr_questions.extra_instructions line uses a
different spacing style than the surrounding keys in the same section. This creates unnecessary
formatting drift in pr_agent/settings TOML defaults.

Code

pr_agent/settings/configuration.toml[132]
+extra_instructions = ""

Evidence
Compliance ID 14 requires preserving TOML formatting in pr_agent/settings/*.toml. In the
[pr_questions] section, existing keys use key=value style (no spaces), while the newly added
extra_instructions = "" introduces a different formatting style within the same section.
AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments
pr_agent/settings/configuration.toml[129-133]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`pr_questions.extra_instructions` is added with inconsistent spacing around `=` compared to adjacent keys in the same TOML section, creating avoidable formatting drift.
## Issue Context
This file is under `pr_agent/settings/`, where formatting/ordering should be preserved as much as possible to reduce noisy diffs.
## Fix Focus Areas
- pr_agent/settings/configuration.toml[129-133]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

5. Test Jinja env mismatch 🐞 Bug ⚙ Maintainability ⭐ New

Description

_render_jinja_template() enables Jinja autoescape=True, but production prompt rendering does
not, so rendered prompts can differ when extra_instructions contains escapable characters (e.g.,
<, &, quotes). This reduces fidelity of the new prompt-rendering tests because they may pass
while production would render different system prompt text.

Code

tests/unittest/test_pr_questions_helpers.py[R21-25]

+def _render_jinja_template(template: str, variables: dict) -> str:
+    from jinja2 import Environment, StrictUndefined
+
+    environment = Environment(undefined=StrictUndefined, autoescape=True)
+    return environment.from_string(template).render(variables)

Evidence
The unit test helper enables autoescaping, while the runtime code paths render the same templates
with a non-autoescaping Jinja environment; therefore, extra_instructions can be escaped in tests
but not in production.
tests/unittest/test_pr_questions_helpers.py[21-26]
pr_agent/tools/pr_questions.py[105-110]
pr_agent/tools/pr_line_questions.py[152-158]
pr_agent/algo/token_handler.py[88-97]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
The prompt-rendering tests use a Jinja2 environment with `autoescape=True`, while production uses `Environment(undefined=StrictUndefined)` (default `autoescape=False`). This means tests can render `extra_instructions` differently than the actual runtime prompt.

Concrete example: if `extra_instructions` contains `<b>no</b>`, tests render `&lt;b&gt;no&lt;/b&gt;` while production renders the literal `<b>no</b>`.

### Issue Context
Production prompt rendering happens in `PRQuestions._get_prediction`, `PR_LineQuestions._get_prediction`, and in `TokenHandler._get_system_user_tokens`, all creating `Environment(undefined=StrictUndefined)` without autoescape.

### Fix Focus Areas
- tests/unittest/test_pr_questions_helpers.py[21-26]
- pr_agent/tools/pr_questions.py[105-110]
- pr_agent/tools/pr_line_questions.py[152-158]
- pr_agent/algo/token_handler.py[88-94]

### Suggested fix
Update `_render_jinja_template()` in the tests to match production:
- Remove `autoescape=True` (or set `autoescape=False`).

(Alternatively, if you intentionally want autoescaping everywhere, then change production environments consistently and update any expectations accordingly.)

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Author self-review: I have reviewed the code review findings, and addressed the relevant ones.

Previous review results

Review updated until commit bea693b

Results up to commit a0ba4f4

🐞 Bugs (2) 📘 Rule violations (1) 📎 Requirement gaps (1) 📜 Skill insights (0)

Context used

✅ Tickets: 🎫 Support "extra_instructions" for /ask

1. Line prompt forces answering always 📎 Requirement gap ⛨ Security ⭐ New

Description

The line-level /ask system prompt still includes unconditional “must always answer” language that
can conflict with extra_instructions intended to require refusals for disallowed questions (e.g.,
PR rating requests). This direct prompt conflict can undermine policy enforcement by causing the
model to answer when it should refuse.

Code

pr_agent/settings/pr_line_questions_prompts.toml[R8-15]

+{%- if extra_instructions %}
+
+
+Extra instructions from the user:
+======
+{{ extra_instructions }}
+======
+{%- endif %}

Evidence

PR Compliance ID 6 requires extra_instructions to steer behavior, including reliably refusing
disallowed questions. The cited /ask system prompt includes an extra_instructions section, but
it is preceded by an unconditional requirement to always answer / not avoid answering, which creates
a direct conflict with refusal-oriented extra_instructions and can prevent refusal behavior from
being applied as required.

Add support for extra_instructions in /ask command
pr_agent/settings/pr_line_questions_prompts.toml[4-16]
pr_agent/settings/pr_questions_prompts.toml[4-15]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The `/ask` system prompt (including the line-level prompt) contains unconditional language requiring the model to always answer questions / not avoid answering, which can contradict `extra_instructions` that are meant to restrict behavior and require refusal for disallowed question types.

## Issue Context
PR Compliance ID 6 expects `extra_instructions` to be enforceable and to steer behavior, including refusing disallowed requests (e.g., “rate this PR 1–10”) with an explanation. Because the current `/ask` prompt includes “must answer” instructions alongside an `extra_instructions` block, the unconditional “always answer” requirement can override or conflict with refusal requirements, undermining compliance for line-level questions.

## Fix Focus Areas
- pr_agent/settings/pr_line_questions_prompts.toml[4-16]
- pr_agent/settings/pr_questions_prompts.toml[4-15]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

2. Extra instructions logged 🐞 Bug ⛨ Security ⭐ New

Description

PRQuestions.run() logs the entire pr_questions config dict; with the new
pr_questions.extra_instructions field, any text configured there will be included in debug log
artifacts.

Code

pr_agent/settings/configuration.toml[132]
+extra_instructions = ""

Evidence
The PR adds extra_instructions under [pr_questions], and the /ask implementation logs the entire
pr_questions section as debug artifacts, which will now include that new field's value.
pr_agent/settings/configuration.toml[129-133]
pr_agent/tools/pr_questions.py[54-58]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`PRQuestions.run()` logs `dict(get_settings().pr_questions)` as an artifact. After this PR, that dict includes `extra_instructions`, so any sensitive internal guidance placed there will be written to logs.

### Issue Context
This PR introduces `pr_questions.extra_instructions` to support policy-like instructions for `/ask`. Those instructions may be more sensitive than typical boolean/tuning settings.

### Fix Focus Areas
- pr_agent/tools/pr_questions.py[54-58]
- pr_agent/settings/configuration.toml[129-133]

### Implementation notes
- When building `relevant_configs`, omit `extra_instructions` entirely, or replace its value with a redacted form (e.g., `"<redacted>"` or `"<redacted: N chars>"`).
- Keep the rest of the config dump intact for debugging.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

3. No token budgeting line-ask 🐞 Bug ☼ Reliability ⭐ New

Description

PR_LineQuestions renders and sends prompts directly to the model without any token-budget
enforcement; with the newly-added extra_instructions block, long instructions can push the request
over the model context limit and fail the AI call.

Code

pr_agent/tools/pr_line_questions.py[R38-39]

+            "conversation_history": "",
+            "extra_instructions": get_settings().pr_questions.extra_instructions,

Evidence
The PR adds extra_instructions into the line-level system prompt and passes it via self.vars,
but PR_LineQuestions._get_prediction() calls the model with fully-rendered prompts and does not
use any token-budgeting/clipping logic before chat_completion.
pr_agent/settings/pr_line_questions_prompts.toml[8-16]
pr_agent/tools/pr_line_questions.py[31-45]
pr_agent/tools/pr_line_questions.py[152-167]
pr_agent/tools/pr_questions.py[40-43]
pr_agent/algo/pr_processing.py[67-96]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
`PR_LineQuestions._get_prediction()` sends `system_prompt`/`user_prompt` to `chat_completion` without any clipping/token budgeting. This PR adds `extra_instructions` into the line-question system prompt, increasing the chance the combined prompt exceeds the model context window.

### Issue Context
Unlike `PRQuestions`, the line-level flow does not use `get_pr_diff(...)` or any other pruning step before calling the model.

### Fix Focus Areas
- pr_agent/tools/pr_line_questions.py[31-45]
- pr_agent/tools/pr_line_questions.py[152-167]
- pr_agent/settings/pr_line_questions_prompts.toml[1-20]

### Implementation notes
Implement a pre-flight token budget check in `PR_LineQuestions._get_prediction()` (or just before it calls `chat_completion`):
- Count tokens for `system_prompt` + `user_prompt`.
- If over limit, clip least-critical fields first (e.g., `conversation_history`, then `full_hunk`, and/or cap `extra_instructions` length) using existing clipping utilities.
- Ensure the truncation is deterministic and logged (without dumping full sensitive content).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

4. extra_instructions TOML formatting inconsistent 📘 Rule violation ⚙ Maintainability

Description

In pr_agent/settings/configuration.toml the new pr_questions.extra_instructions line uses a
different spacing style than the surrounding keys in the same section. This creates unnecessary
formatting drift in pr_agent/settings TOML defaults.

Code

pr_agent/settings/configuration.toml[132]
+extra_instructions = ""

Evidence
Compliance ID 14 requires preserving TOML formatting in pr_agent/settings/*.toml. In the
[pr_questions] section, existing keys use key=value style (no spaces), while the newly added
extra_instructions = "" introduces a different formatting style within the same section.
AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments: AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments
pr_agent/settings/configuration.toml[129-133]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`pr_questions.extra_instructions` is added with inconsistent spacing around `=` compared to adjacent keys in the same TOML section, creating avoidable formatting drift.
## Issue Context
This file is under `pr_agent/settings/`, where formatting/ordering should be preserved as much as possible to reduce noisy diffs.
## Fix Focus Areas
- pr_agent/settings/configuration.toml[129-133]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Results up to commit 009ee19

🐞 Bugs (0) 📘 Rule violations (1) 📎 Requirement gaps (0) 📜 Skill insights (0)

Context used

✅ Tickets: 🎫 Support "extra_instructions" for /ask

1. extra_instructions TOML formatting inconsistent 📘 Rule violation ⚙ Maintainability

Description

In pr_agent/settings/configuration.toml the new pr_questions.extra_instructions line uses a
different spacing style than the surrounding keys in the same section. This creates unnecessary
formatting drift in pr_agent/settings TOML defaults.

Code

pr_agent/settings/configuration.toml[132]
+extra_instructions = ""

Evidence
Compliance ID 14 requires preserving TOML formatting in pr_agent/settings/*.toml. In the
[pr_questions] section, existing keys use key=value style (no spaces), while the newly added
extra_instructions = "" introduces a different formatting style within the same section.
AGENTS.md: Preserve pr_agent/settings TOML Prompt/Default Formatting, Section Order, and Comments
pr_agent/settings/configuration.toml[129-133]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`pr_questions.extra_instructions` is added with inconsistent spacing around `=` compared to adjacent keys in the same TOML section, creating avoidable formatting drift.

## Issue Context
This file is under `pr_agent/settings/`, where formatting/ordering should be preserved as much as possible to reduce noisy diffs.

## Fix Focus Areas
- pr_agent/settings/configuration.toml[129-133]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

rosspeili · 2026-06-29T12:19:37Z

I kept extra_instructions = "" because that's how every other extra_instructions line is written in configuration.toml. Happy to change we need section-local formatting instead. Maybe what's worth a follow-up is adding an /ask example to additional_configurations.md if needed in the same pr?

+            "pr_questions.extra_instructions",
+            "Do not answer questions that ask to rate PR quality.",
+        )
+        environment = Environment(undefined=StrictUndefined)


+        from jinja2 import Environment, StrictUndefined
+
+        extra_instructions_settings.set("pr_questions.extra_instructions", "")
+        environment = Environment(undefined=StrictUndefined)


+            "pr_questions.extra_instructions",
+            "Do not answer questions that ask to rate PR quality.",
+        )
+        environment = Environment(undefined=StrictUndefined)


rosspeili · 2026-06-29T16:00:41Z

Fixed the CodeQL alerts, the new prompt tests now render Jinja templates with autoescape=True. These are plain-text LLM prompts, so behavior is unchanged, also no html involved, these are local tests, still updated them to pass ci.

qodo-free-for-open-source-projects · 2026-06-29T16:04:20Z

Code review by qodo was updated up to the latest commit a0ba4f4

qodo-free-for-open-source-projects · 2026-06-29T16:30:19Z

Code review by qodo was updated up to the latest commit bea693b

feat(ask): add extra_instructions support for /ask (The-PR-Agent#2473)

009ee19

github-actions Bot added the feature 💡 label Jun 29, 2026

rosspeili mentioned this pull request Jun 29, 2026

Support "extra_instructions" for /ask #2473

Open

github-advanced-security AI found potential problems Jun 29, 2026

View reviewed changes

test(ask): enable jinja autoescape in prompt rendering tests

a0ba4f4

qodo-free-for-open-source-projects Bot reviewed Jun 29, 2026

View reviewed changes

Comment thread pr_agent/settings/pr_line_questions_prompts.toml

fix(ask): clarify extra_instructions override base prompt guidance

bea693b

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

feat(ask): add extra_instructions support for /ask (#2473)#2482

feat(ask): add extra_instructions support for /ask (#2473)#2482
rosspeili wants to merge 3 commits into
The-PR-Agent:mainfrom
rosspeili:feature/2473-extra-instructions-for-ask

rosspeili commented Jun 29, 2026

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026 •

edited

Loading

Uh oh!

rosspeili commented Jun 29, 2026 •

edited

Loading

Uh oh!

rosspeili commented Jun 29, 2026

Uh oh!

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

rosspeili commented Jun 29, 2026

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

PR Summary by Qodo

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review by Qodo

Previous review results

Uh oh!

rosspeili commented Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rosspeili commented Jun 29, 2026

Uh oh!

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

Uh oh!

qodo-free-for-open-source-projects Bot commented Jun 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

qodo-free-for-open-source-projects Bot commented Jun 29, 2026 •

edited

Loading

rosspeili commented Jun 29, 2026 •

edited

Loading