CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

26.3 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 1 flagged

Securityscore 75 · 1 finding

HIGH"Ignore previous instructions" command embedded in the skillSS-SKILL-INJECT-IGNORE-01 · Prompt injection · CHANGELOG.md

HIGH — when it fires on hostile content the impact is full system-prompt override.

Why it matters

The text {match} is the classic direct prompt-injection phrasing. Placed in a skill body that the agent reads as trusted instructions, it tries to make the agent abandon its prior rules and follow whatever comes next — a full system-prompt override.

The exact value spotted

excerptCHANGELOG.md· markdown

1116- **orchestrating-swarms/references/context-carry-forward.md**: decision table for long orch

… (108 chars elided on L1116)

1117- **writing/references/pr-descriptions.md**: PR/MR description style guide. Sizing matrix (1

… (108 chars elided on L1117)

1118- **agent-native-architecture/references/dynamic-context-injection.md**: new Trust Levels pr

… (108 chars elided on L1118)

1119

1120### Changed

Occurrences

1 occurrence · at L1118

How to fix

Remove the override phrase, or rephrase the passage so it does not instruct the agent to discard its rules.

Delete the ignore/disregard/forget … previous instructions sentence.
If this is jailbreak-research or tutorial content, move the example into a clearly fenced, non-instruction block and label it as a quoted sample.

Framework references

OWASPLLM01 ATLASAML.T0051

Trace & refs

ruleSS-SKILL-INJECT-IGNORE-01sha256230b1f942b385483…rubric 365aacaView on GitHub

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

distillery/generated-skills/md-docs/SKILL.md · 1 file

distillery/generated-skills/md-docs/SKILL.md3.4 KB · Markdown

name: md-docs

description: >-

Manages project documentation: AGENTS.md, README.md, and CONTRIBUTING.md.

Use when asked to "update README", "update agents", "init agents",

"create AGENTS.md", "update AGENTS.md", "update CONTRIBUTING",

"update context files", or "init context". Not for general markdown editing.

Markdown Documentation

Manage project documentation by verifying against actual codebase state. Emphasize verification over blind generation — analyze structure, files, and patterns before writing.

Portability

AGENTS.md is the universal context file (works with Claude Code, Codex, Kilocode). If the project uses CLAUDE.md, treat it as a symlink to AGENTS.md or migrate content into AGENTS.md and create the symlink:

# If CLAUDE.md exists and AGENTS.md doesn't
mv CLAUDE.md AGENTS.md && ln -sf AGENTS.md CLAUDE.md

When this skill references "context files", it means AGENTS.md (and CLAUDE.md if present as symlink).

Workflows

Update Context Files

Verify and fix AGENTS.md against the actual codebase. See references/update-agents.md for the full verification workflow.

Read existing AGENTS.md, extract verifiable claims (paths, commands, structure, tooling)
Verify each claim against codebase (ls, cat package.json, cat pyproject.toml, etc.)
Fix discrepancies: outdated paths, wrong commands, missing sections, stale structure
Discover undocumented patterns (scripts, build tools, test frameworks not yet documented)
Report changes

Update README

Generate or refresh README.md from project metadata and structure. See references/update-readme.md for section templates and language-specific patterns.

Detect language/stack from config files (package.json, pyproject.toml, composer.json)
Extract metadata: name, version, description, license, scripts
If README exists and --preserve: keep custom sections (About, Features), regenerate standard sections (Install, Usage)
Generate sections appropriate to project type (library vs application)
Report changes

Update CONTRIBUTING

Update existing CONTRIBUTING.md only — never auto-create. See references/update-contributing.md.

Initialize Context

Create AGENTS.md from scratch for projects without documentation. See references/init-agents.md.

Analyze project: language, framework, structure, build/test tools
Generate terse, expert-to-expert context sections
Write AGENTS.md, create CLAUDE.md symlink

Arguments

All workflows support:

--dry-run: preview changes without writing
--preserve: keep existing structure, fix inaccuracies only
--minimal: quick pass, high-level structure only
--thorough: deep analysis of all files

Backup Handling

Before overwriting, back up existing files:

cp AGENTS.md AGENTS.md.backup
cp README.md README.md.backup

Never delete backups automatically.

Writing Style

Terse: omit needless words, lead with the answer
Imperative: "Build the project" not "The project is built"
Expert-to-expert: skip basic explanations, assume competence
Scannable: headings, lists, code blocks
Accurate: verify every command and path against codebase
Sentence case headings, no emoji headers

Report Format

After every operation, display a summary:

✓ Updated AGENTS.md
  - Fixed build command
  - Added new directory to structure

✓ Updated README.md
  - Added installation section
  - Updated badges

⊘ CONTRIBUTING.md not found (skipped)

md-docs— agent skill

Is md-docs safe to install?

Markdown Documentation

Portability

Workflows

Update Context Files

Update README

Update CONTRIBUTING

Initialize Context

Arguments

Backup Handling

Writing Style

Report Format

Audit the pieces. Scan the whole. Decide.