Skip to content

feat: agent-ready — installable SKILL.md skills, AGENTS.md, coding-ag… - #926

Merged
0xallam merged 8 commits into
mainfrom
devin/1785210884-agent-skills
Aug 6, 2026
Merged

feat: agent-ready — installable SKILL.md skills, AGENTS.md, coding-ag…#926
0xallam merged 8 commits into
mainfrom
devin/1785210884-agent-skills

Conversation

@bearsyankees

Copy link
Copy Markdown
Collaborator

…ents docs

@greptile-apps

greptile-apps Bot commented Jul 28, 2026

Copy link
Copy Markdown
Contributor

Greptile Summary

Adds installable agent skills and supporting documentation for running Strix, remediating findings, using the cloud API, and configuring CI.

  • Adds four SKILL.md workflows for pentesting, cloud API usage, finding remediation, and CI setup.
  • Adds AGENTS.md and a coding-agent integration guide.
  • Updates the README and documentation navigation to advertise the agent workflows.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename Overview
skills/strix-ci-setup/SKILL.md Documents self-hosted and managed CI integration, including validated diff-base resolution and a fail-closed completion-status gate.
skills/strix-fix-findings/SKILL.md Documents remediation and verification workflows while avoiding unsafe upstream and single-commit diff-base fallbacks.
skills/strix-pentest/SKILL.md Adds agent instructions for self-hosted and managed scans, result interpretation, and truncated-run handling.
skills/strix-cloud-api/SKILL.md Adds a managed-platform API workflow covering assets, scans, findings, reports, PR reviews, schedules, and webhooks.
docs/integrations/coding-agents.mdx Adds user-facing documentation for installing and selecting the new coding-agent skills.
AGENTS.md Adds a concise guide for agents using or contributing to Strix.

Reviews (5): Last reviewed commit: "docs(skills): enforce scan completion in..." | Re-trigger Greptile

Comment thread skills/strix-ci-setup/SKILL.md
Comment thread skills/strix-fix-findings/SKILL.md Outdated
- New strix-cloud-api skill: token/scopes, register assets, launch+poll
  scans, read findings, SARIF/report export, PR reviews, schedules, webhooks
- strix-pentest: unbiased OSS-vs-Cloud decision guide; both run paths
- strix-fix-findings, strix-ci-setup: cover cloud verify + managed PR reviews
- AGENTS.md/README/coding-agents docs: document both run modes
- CI + verify: resolve the PR/default base branch instead of hard-coding
  origin/main (repos use master/develop; GitHub Actions uses github.base_ref)
- Document the budget/fail-open caveat: exit 0 means clean for what was
  analyzed, not full coverage; size budget to complete and check run.json
@devin-ai-integration

Copy link
Copy Markdown
Contributor

@greptile

Comment thread skills/strix-fix-findings/SKILL.md Outdated
Comment thread skills/strix-ci-setup/SKILL.md Outdated
- Drop @{u} fallback that could resolve the current feature branch as its
  own diff base (empty diff -> falsely clean verification)
- Prefer origin/HEAD, then a common default branch that exists, else HEAD~1
- CI: resolve base branch without pipe-masking (parameter expansion + ref
  verification) so a failed lookup can't silently produce origin/
@devin-ai-integration

Copy link
Copy Markdown
Contributor

@greptile

Comment thread skills/strix-ci-setup/SKILL.md Outdated
Comment thread skills/strix-fix-findings/SKILL.md Outdated
Never fall back to HEAD~1 - on a multi-commit branch it scans only the
last commit and earlier changes can pass unscanned. Require an explicit
base (or the diff-base-free --instruction verification) instead.
@devin-ai-integration

Copy link
Copy Markdown
Contributor

@greptile

Comment thread skills/strix-fix-findings/SKILL.md
…gin/HEAD

- CI workflow: add explicit step failing the build unless run.json status
  is 'completed' (budget-truncated runs are 'stopped' yet exit 0)
- fix-findings: verify origin/HEAD's target exists before using it as the
  diff base so a dangling symbolic ref falls through to the fallbacks
@devin-ai-integration

Copy link
Copy Markdown
Contributor

@greptile

5hy7xz92nd-oss

This comment was marked as spam.

@bearsyankees
bearsyankees requested a review from 0xallam July 28, 2026 05:17
@devin-ai-integration

Copy link
Copy Markdown
Contributor

Reviewed the skills against the actual implementations and pushed fixes in bd21342.

Wrong against strix-app's src/lib/api-v1/openapi.json (an agent following them got 4xx):

  • domain registration used "asset_type":"web" — the enum is web_app | api | attack_surface
  • repository registration wrapped the body in {"repositories":[...]} — the route accepts one object or a bare array, so the wrapped form failed with full_name is required for every repository
  • report download used Accept: application/pdf — format/file type are query params (?format=technical&type=pdf|docx); Accept is ignored

Accuracy/caveats tightened:

  • report downloads are gated on the Enterprise compliance_reports feature (DOCX doubly so), but PDF/DOCX reports were sold as a generic managed-cloud benefit in AGENTS.md, the docs page, and strix-pentest
  • the CI "fail unless status == completed" gate only catches the hard budget stop (BudgetExceededErrorstopped); a run that wraps up on a budget warning still calls finish_scan and records completed with partial coverage, so the note now says to also compare run.json's llm_usage.cost against --max-budget
  • jq 'sort_by(.severity)' sorted alphabetically; replaced with an explicit severity order, plus the cloud severity/status enums

Everything else checked out: npx skills add discovers all four skills with valid frontmatter, CLI flags/exit codes/artifact names match strix/interface/main.py and strix/report/writer.py, and every linked URL resolves.

@0xallam
0xallam merged commit 2a9ab1d into main Aug 6, 2026
@0xallam
0xallam deleted the devin/1785210884-agent-skills branch August 6, 2026 13:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants